Clasificación Aviar en Ibagué mediante Deep Learning: Enfoque en Precisión y Eficiencia

El proceso de clasificación aviar en Ibagué, mediante técnicas de Deep Learning, reviste una importancia fundamental en la conservación y el estudio de la biodiversidad de la región. Ibagué, ubicada en una zona geográfica rica en diversidad de aves, se enfrenta a desafíos significativos en la identificación y monitoreo de estas especies. La implementación de un sistema multiclase de identificación aviar no solo facilita la tarea de catalogación para investigadores y conservacionistas, sino que también proporciona una herramienta eficaz para la evaluación del estado de las poblaciones aviares y la detección temprana de posibles amenazas, como la pérdida de hábitat o el cambio climático. Con un enfoque en la precisión y eficiencia del proceso de clasificación, este sistema puede contribuir de manera significativa a la gestión sostenible de los recursos naturales y la protección del medio ambiente en la región de Ibagué.

Librerias

A continuación se relacionana algunas de las bibliotecas y módulos de Python que se utilizan comúnmente en el desarrollo de aplicaciones de aprendizaje automático, procesamiento de imágenes y que se utilizaran en este proyecto:

1. Keras: Es una biblioteca de redes neuronales de código abierto escrita en Python que facilita la creación y entrenamiento de modelos de aprendizaje profundo. Proporciona una interfaz simple y consistente para construir y entrenar modelos de redes neuronales.

2. NumPy: Es una biblioteca fundamental para computación numérica en Python. Se utiliza para realizar operaciones matemáticas en matrices y matrices multidimensionales, lo que es esencial para el procesamiento de datos en el aprendizaje automático.

3. OpenCV (cv2): OpenCV (Open Source Computer Vision Library) es una biblioteca de código abierto que se utiliza para el procesamiento de imágenes y visión por computadora. Proporciona una amplia gama de funciones y algoritmos para tareas como manipulación de imágenes, detección de características, reconocimiento de objetos, seguimiento de objetos, entre otros.

4. Matplotlib: Es una biblioteca de trazado en 2D de Python que produce figuras de calidad de publicación en una variedad de formatos y entornos. Se utiliza para visualizar datos y resultados, incluidas imágenes y gráficos.

5. TensorFlow: Es una biblioteca de código abierto desarrollada por Google para el aprendizaje automático y la inteligencia artificial. TensorFlow proporciona un ecosistema completo para construir y entrenar modelos de aprendizaje profundo, incluidas API de alto nivel como Keras.

6. Scikit-learn: Es una biblioteca de aprendizaje automático de código abierto que proporciona herramientas simples y eficientes para el análisis predictivo y la minería de datos. Incluye una variedad de algoritmos de aprendizaje supervisado y no supervisado, así como herramientas para preprocesamiento de datos, evaluación de modelos y selección de características.

En resumen, estas bibliotecas son componentes clave en el desarrollo de aplicaciones de aprendizaje automático y procesamiento de imágenes, proporcionando herramientas y funciones para construir, entrenar, evaluar y visualizar modelos de manera eficiente.

from keras.models import Sequential, Model
from keras.layers import Conv2D, MaxPool2D, Dense, Flatten, Dropout, BatchNormalization, Input
from keras.optimizers import Adam
from keras.callbacks import TensorBoard, ModelCheckpoint
from keras.utils import to_categorical
import os
import numpy as np
from keras.preprocessing import image
from keras.applications.imagenet_utils import preprocess_input, decode_predictions
from keras.applications.vgg16 import VGG16
from tensorflow.keras.preprocessing.image import ImageDataGenerator 
from tensorflow.keras.models import load_model
from sklearn.utils import shuffle
from sklearn.model_selection import train_test_split
import cv2
import matplotlib.pyplot as plt
%matplotlib inline

Configuración de parámetros

Es crucial establecer los parámetros específicos, como el tamaño de las imágenes de entrada (width_shape y height_shape), el número de clases a predecir (num_classes), así como los hiperparámetros de entrenamiento, como el número de épocas (epochs) y el tamaño del lote (batch_size), antes de iniciar el proceso de entrenamiento de un modelo de aprendizaje profundo para la clasificación de imágenes. Estos parámetros tienen un impacto significativo en la calidad y eficiencia del modelo resultante. Por ejemplo, el tamaño de las imágenes de entrada afecta la capacidad del modelo para capturar detalles importantes de las imágenes, mientras que el número de clases determina la complejidad de la tarea de clasificación. Además, la configuración adecuada de los hiperparámetros de entrenamiento, como el número de épocas y el tamaño del lote, es esencial para garantizar un entrenamiento efectivo y eficiente del modelo, evitando el sobreajuste o el subajuste. Por lo tanto, la correcta especificación de estos parámetros es fundamental para lograr resultados precisos y confiables en el proceso de clasificación de imágenes.

width_shape = 224
height_shape = 224
num_classes = 18
epochs = 50
batch_size = 32

Este código define algunas variables que son comunes al entrenar modelos de aprendizaje profundo para clasificación de imágenes:

• width_shape y height_shape: Estas variables representan el ancho y la altura de las imágenes de entrada al modelo. En este caso, las imágenes se están redimensionando a una forma cuadrada de 224x224 píxeles. Es común redimensionar las imágenes de entrada a un tamaño específico antes de alimentarlas al modelo.

• num_classes: Esta variable indica el número de clases en el problema de clasificación. En este caso, el modelo se entrenará para clasificar imágenes en 10 clases diferentes.

• epochs: Este parámetro especifica el número de épocas o iteraciones completas sobre el conjunto de datos durante el entrenamiento del modelo. Cada época implica pasar por todo el conjunto de datos una vez hacia adelante y hacia atrás a través de la red neuronal.

• batch_size: Este parámetro indica el tamaño del lote de datos que se utilizará en cada iteración de entrenamiento. Durante el entrenamiento, los datos se dividen en lotes y se procesan en paralelo, lo que puede acelerar el proceso de entrenamiento y hacerlo más eficiente en términos de uso de memoria. Un valor típico para batch_size es 32, lo que significa que se procesarán 32 imágenes a la vez durante cada iteración de entrenamiento.

Path de dataset

Estos son los directorios donde se almacenan los datos de entrenamiento y validación respectivamente para el modelo. Es una práctica común dividir el conjunto de datos en un conjunto de entrenamiento y un conjunto de validación para evaluar el rendimiento del modelo durante el entrenamiento.

train_data_dir = 'dataset/train'  
validation_data_dir = 'dataset/valid'

• train_data_dir: Este directorio contiene las imágenes de entrenamiento. Por lo general, se espera que contenga subdirectorios separados para cada clase (cada tipo de ave) de imagen, donde las imágenes de cada clase se agrupan juntas.

• validation_data_dir: Este directorio contiene las imágenes de validación, que se utilizan para evaluar el rendimiento del modelo en un conjunto de datos independiente durante el entrenamiento. Al igual que con el directorio de entrenamiento, se espera que este directorio contenga subdirectorios separados para cada clase de imagen.

La estructura de directorios esperada por los generadores de datos de Keras o TensorFlow para tareas de clasificación de imágenes suele ser la siguiente:

dataset/
├── train/
│   ├── class1/
│   │   ├── image1.jpg
│   │   ├── image2.jpg
│   │   └── ...
│   ├── class2/
│   │   ├── image1.jpg
│   │   ├── image2.jpg
│   │   └── ...
│   └── ...
└── valid/
    ├── class1/
    │   ├── image1.jpg
    │   ├── image2.jpg
    │   └── ...
    ├── class2/
    │   ├── image1.jpg
    │   ├── image2.jpg
    │   └── ...
    └── ...

En esta estructura:

• El directorio train contiene subdirectorios separados para cada clase de imagen.
• Cada subdirectorio de clase (class1, class2, etc.) contiene las imágenes correspondientes a esa clase.
• El directorio valid tiene una estructura similar al directorio de entrenamiento y se utiliza para almacenar las imágenes de validación.

Es importante seguir esta estructura de directorios para que los generadores de datos de Keras o TensorFlow puedan encontrar y cargar correctamente las imágenes durante el entrenamiento y la validación del modelo. Si los datos no están organizados de esta manera, los generadores de datos pueden tener dificultades para encontrar las imágenes y el proceso de entrenamiento puede fallar o producir resultados incorrectos.

Generador de imágenes para las imagenes de train y valid

Los generadores de imágenes son fundamentales para cargar y procesar los datos de entrenamiento y validación de manera eficiente durante el entrenamiento del modelo de clasificación de imágenes. Además, aplican aumentos de datos en tiempo real durante el entrenamiento, lo que ayuda a mejorar la capacidad del modelo para generalizar a nuevos datos.

# Definir el generador de imágenes para el conjunto de entrenamiento con aumentos de datos
train_datagen = ImageDataGenerator(  
    rotation_range=20,               # Rango de grados para rotación aleatoria
    zoom_range=0.2,                  # Rango de zoom aleatorio
    width_shift_range=0.1,           # Rango de desplazamiento horizontal aleatorio
    height_shift_range=0.1,          # Rango de desplazamiento vertical aleatorio
    horizontal_flip=True,            # Volteo horizontal aleatorio
    vertical_flip=False,             # No se aplica volteo vertical
    preprocessing_function=preprocess_input)  # Función de preprocesamiento

# Definir el generador de imágenes para el conjunto de validación con los mismos aumentos de datos
valid_datagen = ImageDataGenerator(    
    rotation_range=20,
    zoom_range=0.2,
    width_shift_range=0.1,
    height_shift_range=0.1,
    horizontal_flip=True,
    vertical_flip=False,
    preprocessing_function=preprocess_input)

# Crear un generador de lotes de imágenes para el conjunto de entrenamiento
train_generator = train_datagen.flow_from_directory(  
    train_data_dir,                  # Directorio que contiene las imágenes de entrenamiento
    target_size=(width_shape, height_shape),  # Tamaño al que se redimensionarán las imágenes
    batch_size=batch_size,           # Tamaño del lote
    class_mode='categorical')        # Modo de clasificación para imágenes categóricas

# Crear un generador de lotes de imágenes para el conjunto de validación
validation_generator = valid_datagen.flow_from_directory(  
    validation_data_dir,             # Directorio que contiene las imágenes de validación
    target_size=(width_shape, height_shape),  # Tamaño al que se redimensionarán las imágenes
    batch_size=batch_size,           # Tamaño del lote
    class_mode='categorical')        # Modo de clasificación para imágenes categóricas

Found 834 images belonging to 5 classes.
Found 25 images belonging to 5 classes.

Entrenamiento de modelo VGG16

A continuación, se configura y entrena un modelo de red neuronal convolucional (CNN) utilizando la arquitectura VGG16 preentrenada para la clasificación de imágenes. Primero, se establecen variables para el número de muestras de entrenamiento y validación. Luego, se define la estructura de entrada de la red neuronal y se carga el modelo VGG16 preentrenado con sus pesos ajustados a partir del conjunto de datos ImageNet. Se añade una capa densa adicional para adaptar la salida a la cantidad de clases en el problema de clasificación y se congela el resto de las capas para evitar que se modifiquen durante el entrenamiento. Posteriormente, el modelo se compila con una función de pérdida y un optimizador específicos, y se muestra un resumen de la arquitectura del modelo. Finalmente, se lleva a cabo el entrenamiento del modelo utilizando generadores de datos para el conjunto de entrenamiento y validación, con la configuración de épocas y pasos definida previamente. Durante el entrenamiento, se actualizan únicamente los pesos de la capa densa añadida, mientras que los pesos de las capas preentrenadas permanecen constantes debido a la congelación previa.

Prueba 1

Regularización L2

Se agrega regularización L2 a la capa densa utilizando el parámetro kernel_regularizer='l2'. La regularización L2 es una técnica comúnmente utilizada para prevenir el sobreajuste en modelos de aprendizaje automático al penalizar los pesos grandes en la función de pérdida. Esto puede ayudar a mejorar la generalización del modelo, especialmente cuando se tienen conjuntos de datos pequeños o complejos.

Entrenamiento

# Definir el número de muestras de entrenamiento y validación
nb_train_samples = 2805
nb_validation_samples = 90

# Definir la entrada de la red neuronal con el tamaño de las imágenes
image_input = Input(shape=(width_shape, height_shape, 3))

# Cargar el modelo VGG16 preentrenado con pesos ajustados desde ImageNet
model = VGG16(input_tensor=image_input, include_top=True, weights='imagenet')

# Obtener la salida de la penúltima capa densa del modelo VGG16 (fc2)
last_layer = model.get_layer('fc2').output

# Añadir una nueva capa densa al final del modelo para la clasificación multiclase con regularización L2 (Evita sobreajuste)
out = Dense(num_classes, activation='softmax', kernel_regularizer='l2', name='output')(last_layer)

# Crear un nuevo modelo personalizado que toma la entrada de la imagen y produce la salida clasificada
custom_vgg_model = Model(image_input, out)

# Congelar todas las capas del modelo, excepto la capa densa añadida
for layer in custom_vgg_model.layers[:-1]:
    layer.trainable = False

# Compilar el modelo con una función de pérdida, optimizador y métricas especificadas
custom_vgg_model.compile(loss='categorical_crossentropy', optimizer=Adam(learning_rate=0.0001), metrics=['accuracy'])


# Mostrar un resumen del modelo que incluye la arquitectura y el número de parámetros
custom_vgg_model.summary()

# Entrenar el modelo utilizando generadores de datos para el conjunto de entrenamiento y validación
model_history = custom_vgg_model.fit(
    train_generator,
    epochs=epochs,
    validation_data=validation_generator,
    steps_per_epoch=nb_train_samples//batch_size,  # Número de pasos por época de entrenamiento
    validation_steps=nb_validation_samples//batch_size)  # Número de pasos por época de validación

Model: "functional_1"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer (InputLayer)        │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ output (Dense)                  │ (None, 18)             │        73,746 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 134,334,290 (512.44 MB)

 Trainable params: 73,746 (288.07 KB)

 Non-trainable params: 134,260,544 (512.16 MB)

Epoch 1/50

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\keras\src\trainers\data_adapters\py_dataset_adapter.py:120: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.
  self._warn_if_super_not_called()

87/87 ━━━━━━━━━━━━━━━━━━━━ 327s 4s/step - accuracy: 0.2574 - loss: 3.3109 - val_accuracy: 0.6719 - val_loss: 1.3040
Epoch 2/50
 1/87 ━━━━━━━━━━━━━━━━━━━━ 4:41 3s/step - accuracy: 0.7500 - loss: 1.1916

C:\Users\Oscar Diaz\anaconda3\Lib\contextlib.py:158: UserWarning: Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches. You may need to use the `.repeat()` function when building your dataset.
  self.gen.throw(typ, value, traceback)

87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 34ms/step - accuracy: 0.7500 - loss: 1.1916 - val_accuracy: 0.8077 - val_loss: 1.1709
Epoch 3/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 315s 4s/step - accuracy: 0.8064 - loss: 1.0603 - val_accuracy: 0.9219 - val_loss: 0.7147
Epoch 4/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 32ms/step - accuracy: 0.7812 - loss: 0.9347 - val_accuracy: 0.8462 - val_loss: 0.8326
Epoch 5/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 315s 4s/step - accuracy: 0.8887 - loss: 0.7424 - val_accuracy: 0.8750 - val_loss: 0.7001
Epoch 6/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 33ms/step - accuracy: 0.9688 - loss: 0.6136 - val_accuracy: 0.7692 - val_loss: 0.8264
Epoch 7/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 311s 4s/step - accuracy: 0.9222 - loss: 0.6504 - val_accuracy: 0.9844 - val_loss: 0.5047
Epoch 8/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 33ms/step - accuracy: 0.8750 - loss: 0.7784 - val_accuracy: 0.9231 - val_loss: 0.6585
Epoch 9/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 306s 3s/step - accuracy: 0.9399 - loss: 0.5582 - val_accuracy: 0.9219 - val_loss: 0.5437
Epoch 10/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.9375 - loss: 0.5939 - val_accuracy: 0.9615 - val_loss: 0.5021
Epoch 11/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 298s 3s/step - accuracy: 0.9530 - loss: 0.5215 - val_accuracy: 0.9062 - val_loss: 0.5525
Epoch 12/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.8750 - loss: 0.5664 - val_accuracy: 1.0000 - val_loss: 0.4143
Epoch 13/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 299s 3s/step - accuracy: 0.9540 - loss: 0.4771 - val_accuracy: 0.9688 - val_loss: 0.4520
Epoch 14/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 1.0000 - loss: 0.3820 - val_accuracy: 0.8846 - val_loss: 0.5025
Epoch 15/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 303s 3s/step - accuracy: 0.9615 - loss: 0.4615 - val_accuracy: 0.9531 - val_loss: 0.4085
Epoch 16/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9375 - loss: 0.4576 - val_accuracy: 0.8846 - val_loss: 0.4775
Epoch 17/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 307s 4s/step - accuracy: 0.9693 - loss: 0.4188 - val_accuracy: 0.9375 - val_loss: 0.4464
Epoch 18/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 32ms/step - accuracy: 0.9688 - loss: 0.4436 - val_accuracy: 1.0000 - val_loss: 0.4011
Epoch 19/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.9695 - loss: 0.4057 - val_accuracy: 0.9062 - val_loss: 0.4726
Epoch 20/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 13s 121ms/step - accuracy: 0.9688 - loss: 0.4776 - val_accuracy: 0.9231 - val_loss: 0.4076
Epoch 21/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 304s 3s/step - accuracy: 0.9769 - loss: 0.3775 - val_accuracy: 0.9531 - val_loss: 0.3875
Epoch 22/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9062 - loss: 0.4468 - val_accuracy: 0.9615 - val_loss: 0.3648
Epoch 23/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 307s 4s/step - accuracy: 0.9741 - loss: 0.3705 - val_accuracy: 0.9844 - val_loss: 0.3266
Epoch 24/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.9375 - loss: 0.4013 - val_accuracy: 0.9231 - val_loss: 0.4140
Epoch 25/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 298s 3s/step - accuracy: 0.9795 - loss: 0.3441 - val_accuracy: 0.9688 - val_loss: 0.3973
Epoch 26/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 1.0000 - loss: 0.2807 - val_accuracy: 1.0000 - val_loss: 0.2925
Epoch 27/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 296s 3s/step - accuracy: 0.9809 - loss: 0.3312 - val_accuracy: 0.9688 - val_loss: 0.3330
Epoch 28/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 36ms/step - accuracy: 1.0000 - loss: 0.2897 - val_accuracy: 0.9231 - val_loss: 0.4058
Epoch 29/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 298s 3s/step - accuracy: 0.9791 - loss: 0.3303 - val_accuracy: 0.9531 - val_loss: 0.3856
Epoch 30/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 1.0000 - loss: 0.3215 - val_accuracy: 0.9615 - val_loss: 0.3279
Epoch 31/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 297s 3s/step - accuracy: 0.9853 - loss: 0.3096 - val_accuracy: 0.9844 - val_loss: 0.3010
Epoch 32/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9688 - loss: 0.3030 - val_accuracy: 0.9615 - val_loss: 0.4724
Epoch 33/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 292s 3s/step - accuracy: 0.9826 - loss: 0.3023 - val_accuracy: 0.9062 - val_loss: 0.3939
Epoch 34/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9688 - loss: 0.3192 - val_accuracy: 1.0000 - val_loss: 0.2482
Epoch 35/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 292s 3s/step - accuracy: 0.9876 - loss: 0.2885 - val_accuracy: 0.9844 - val_loss: 0.2896
Epoch 36/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9688 - loss: 0.2885 - val_accuracy: 1.0000 - val_loss: 0.2804
Epoch 37/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 299s 3s/step - accuracy: 0.9845 - loss: 0.2804 - val_accuracy: 0.9688 - val_loss: 0.2981
Epoch 38/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 1.0000 - loss: 0.2669 - val_accuracy: 0.9615 - val_loss: 0.2888
Epoch 39/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.9828 - loss: 0.2720 - val_accuracy: 0.9844 - val_loss: 0.2701
Epoch 40/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9688 - loss: 0.2661 - val_accuracy: 1.0000 - val_loss: 0.2824
Epoch 41/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 293s 3s/step - accuracy: 0.9883 - loss: 0.2564 - val_accuracy: 0.9844 - val_loss: 0.2613
Epoch 42/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 29ms/step - accuracy: 1.0000 - loss: 0.2392 - val_accuracy: 0.9231 - val_loss: 0.4176
Epoch 43/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.9897 - loss: 0.2477 - val_accuracy: 0.9844 - val_loss: 0.2354
Epoch 44/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 1.0000 - loss: 0.2093 - val_accuracy: 0.9615 - val_loss: 0.3279
Epoch 45/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 291s 3s/step - accuracy: 0.9914 - loss: 0.2405 - val_accuracy: 0.9531 - val_loss: 0.2933
Epoch 46/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 29ms/step - accuracy: 1.0000 - loss: 0.2711 - val_accuracy: 1.0000 - val_loss: 0.1968
Epoch 47/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 291s 3s/step - accuracy: 0.9890 - loss: 0.2319 - val_accuracy: 0.9688 - val_loss: 0.2634
Epoch 48/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 1.0000 - loss: 0.1956 - val_accuracy: 1.0000 - val_loss: 0.2144
Epoch 49/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.9883 - loss: 0.2314 - val_accuracy: 0.9375 - val_loss: 0.3072
Epoch 50/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 1.0000 - loss: 0.1996 - val_accuracy: 1.0000 - val_loss: 0.1915

Grabar modelo en disco

import os

# Nombre base del modelo
model_name = "model_VGG16_v"

# Extensión del archivo
file_extension = ".keras"

# Directorio donde se guardarán los modelos
model_directory = "models/"

# Inicializar contador
counter = 1

# Generar el nombre completo del archivo
file_name = model_name + file_extension

ruta=model_directory + file_name
print(ruta)
# Verificar si el modelo ya está guardado
while os.path.exists(model_directory + file_name):
    
    # Si el archivo existe, agregar un número al final del nombre del modelo
    file_name = f"{model_name}{counter}{file_extension}"
    counter += 1

# Guardar el modelo con el nombre único en el directorio correcto
custom_vgg_model.save(model_directory + file_name)

models/model_VGG16_v.keras

Gráficas de entrenamiento y validación (accuracy - loss)

Esta función plotTraining se encarga de visualizar la evolución de la pérdida (loss) y la precisión (accuracy) durante el entrenamiento y la validación de un modelo de redes neuronales a lo largo de las épocas.

def plotTraining(hist, epochs, typeData):

    # Seleccionar la figura y establecer el tamaño
    # Dependiendo del tipo de datos (loss o accuracy), se elige la figura correspondiente
    
    if typeData=="loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'-r',label='Loss Training')
    if typeData=="accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'-r',label='Accuracy Training')
    if typeData=="val_loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['val_loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'--b',label='Loss Validate')
    if typeData=="val_accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['val_accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'--b',label='Training Validate')
        

    plt.rc('xtick',labelsize=24)
    plt.rc('ytick',labelsize=24)
    plt.rc('legend', fontsize=18) 
    plt.legend()
    plt.xlabel('Number of Epochs',fontsize=24)
    plt.grid(True)

plotTraining(model_history,epochs,"loss")
plotTraining(model_history,epochs,"accuracy")
plotTraining(model_history,epochs,"val_loss")
plotTraining(model_history,epochs,"val_accuracy")

Predicción usando el modelo entrenado

from keras.applications.imagenet_utils import preprocess_input, decode_predictions
from keras.models import load_model

names = ['CYANOCORAX YNCAS', 'PIRANGA RUBRA', 'PITANGUS SULPHURATUS', 'PYROCEPHALUS RUBINUS', 'RUPORNIS MAGNIROSTRIS', 'SICALIS FLAVEOLA', 'THRAUPIS EPISCOPUS', 'TIARIS OLIVACEUS', 'TYRANNUS MELANCHOLICUS', 'ZONOTRICHIA CAPENSIS', 'CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS', 'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS', 'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA', 'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']

# Cargar el modelo
modelt = load_model(model_directory + file_name)
#modelt = custom_vgg_model

# Ruta de la imagen de prueba
imaget_path = "ImagenPrueba.jpg"

# Leer la imagen, cambiar tamaño y preprocesar
imaget=cv2.resize(cv2.imread(imaget_path), (width_shape, height_shape), interpolation = cv2.INTER_AREA)
xt = np.asarray(imaget)
xt=preprocess_input(xt)
xt = np.expand_dims(xt,axis=0)

# Obtener las predicciones del modelo
preds = modelt.predict(xt)

# Obtener la clase predicha y su porcentaje de confianza
predicted_class_index = np.argmax(preds)
predicted_class_name = names[predicted_class_index]
confidence_percentage = preds[0][predicted_class_index] * 100

# Imprimir el resultado
print(f'Clase predicha: {predicted_class_name}')
print(f'Porcentaje de confianza: {confidence_percentage:.2f}%')

# Mostrar la imagen
plt.imshow(cv2.cvtColor(np.asarray(imaget), cv2.COLOR_BGR2RGB))
plt.axis('off')
plt.show()

1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 785ms/step
Clase predicha: EGRETTA THULA
Porcentaje de confianza: 66.31%

Matriz de confusión y métricas de desempeño

from sklearn.metrics import confusion_matrix, f1_score, roc_curve, precision_score, recall_score, accuracy_score, roc_auc_score
from sklearn import metrics
from mlxtend.plotting import plot_confusion_matrix
from keras.models import load_model
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline


names = ['CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS', 'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS', 'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA', 'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']



test_data_dir = 'dataset/test'  

test_datagen = ImageDataGenerator()

test_generator = test_datagen.flow_from_directory(
    test_data_dir,
    target_size=(width_shape, height_shape), 
    batch_size = batch_size,
    class_mode='categorical', 
    shuffle=False)

custom_Model= load_model(model_directory + file_name)

predictions = custom_Model.predict(test_generator)


y_pred = np.argmax(predictions, axis=1)
y_real = test_generator.classes


matc=confusion_matrix(y_real, y_pred)



print(metrics.classification_report(y_real,y_pred, digits = 4))

Found 85 images belonging to 18 classes.

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\keras\src\trainers\data_adapters\py_dataset_adapter.py:120: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.
  self._warn_if_super_not_called()

3/3 ━━━━━━━━━━━━━━━━━━━━ 9s 3s/step
              precision    recall  f1-score   support

           0     1.0000    0.8000    0.8889         5
           1     0.6667    0.8000    0.7273         5
           2     0.8333    1.0000    0.9091         5
           3     1.0000    0.8000    0.8889         5
           4     0.5556    1.0000    0.7143         5
           5     0.8333    1.0000    0.9091         5
           6     0.5000    1.0000    0.6667         5
           7     0.5556    1.0000    0.7143         5
           8     0.8000    0.8000    0.8000         5
           9     0.3333    0.8000    0.4706         5
          10     1.0000    0.4000    0.5714         5
          11     1.0000    0.6000    0.7500         5
          12     0.0000    0.0000    0.0000         5
          13     0.0000    0.0000    0.0000         5
          14     1.0000    0.6000    0.7500         5
          15     1.0000    0.4000    0.5714         5
          16     0.0000    0.0000    0.0000         5
          17     0.0000    0.0000    0.0000         0

    accuracy                         0.6471        85
   macro avg     0.6154    0.6111    0.5740        85
weighted avg     0.6516    0.6471    0.6078        85

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))

# Define el tamaño de la figura y los nombres de las clases
fig, ax = plot_confusion_matrix(conf_mat=matc, figsize=(18,18), class_names=names, show_normed=True)

# Ajusta el tamaño de las letras de los nombres de las clases
ax.set_xticklabels(names, fontsize=12)  # Ajusta el tamaño de las letras en el eje x
ax.set_yticklabels(names, fontsize=12)  # Ajusta el tamaño de las letras en el eje y

# Ajusta automáticamente el diseño de la figura
plt.tight_layout()

# Muestra la figura
plt.show()

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\mlxtend\plotting\plot_confusion_matrix.py:102: RuntimeWarning: invalid value encountered in divide
  normed_conf_mat = conf_mat.astype("float") / total_samples

Prueba 2

Generador de imágenes para las imagenes de train y valid

# Definir el generador de imágenes para el conjunto de entrenamiento con aumentos de datos
train_datagen = ImageDataGenerator(
    rotation_range=20,               
    zoom_range=0.2,                  
    width_shift_range=0.1,           
    height_shift_range=0.1,          
    horizontal_flip=True,            
    vertical_flip=False,             
    preprocessing_function=preprocess_input
)

# Definir el generador de imágenes para el conjunto de validación con los mismos aumentos de datos
valid_datagen = ImageDataGenerator(
    rotation_range=20,
    zoom_range=0.2,
    width_shift_range=0.1,
    height_shift_range=0.1,
    horizontal_flip=True,
    vertical_flip=False,
    preprocessing_function=preprocess_input
)

# Definir el número de conjuntos de entrenamiento y validación
num_splits = 3

# Dividir los datos en múltiples conjuntos de entrenamiento y validación
for i in range(num_splits):
    # Crear un generador de lotes de imágenes para el conjunto de entrenamiento
    train_generator = train_datagen.flow_from_directory(
        train_data_dir,
        target_size=(width_shape, height_shape),
        batch_size=batch_size,
        class_mode='categorical'
    )

    # Crear un generador de lotes de imágenes para el conjunto de validación
    validation_generator = valid_datagen.flow_from_directory(
        validation_data_dir,
        target_size=(width_shape, height_shape),
        batch_size=batch_size,
        class_mode='categorical'
    )

Found 2805 images belonging to 18 classes.
Found 90 images belonging to 18 classes.
Found 2805 images belonging to 18 classes.
Found 90 images belonging to 18 classes.
Found 2805 images belonging to 18 classes.
Found 90 images belonging to 18 classes.

Entrenamiento

from keras.callbacks import LearningRateScheduler
import math

# Definir el número de muestras de entrenamiento y validación
nb_train_samples = 2805
nb_validation_samples = 90

# Definir la entrada de la red neuronal con el tamaño de las imágenes
image_input = Input(shape=(width_shape, height_shape, 3))

# Cargar el modelo VGG16 preentrenado con pesos ajustados desde ImageNet
model = VGG16(input_tensor=image_input, include_top=True, weights='imagenet')

# Obtener la salida de la penúltima capa densa del modelo VGG16 (fc2)
last_layer = model.get_layer('fc2').output

# Añadir una nueva capa densa al final del modelo para la clasificación multiclase con regularización L2 (Evita sobreajuste)
out = Dense(num_classes, activation='softmax', kernel_regularizer='l2', name='output')(last_layer)

# Crear un nuevo modelo personalizado que toma la entrada de la imagen y produce la salida clasificada
custom_vgg_model = Model(image_input, out)

# Congelar todas las capas del modelo, excepto la capa densa añadida
for layer in custom_vgg_model.layers[:-1]:
    layer.trainable = False

# Compilar el modelo con una función de pérdida, optimizador y métricas especificadas
custom_vgg_model.compile(loss='categorical_crossentropy', optimizer=Adam(learning_rate=0.0001), metrics=['accuracy'])

# Definir una función que reduzca la tasa de aprendizaje a la mitad cada 2 épocas
def lr_schedule(epoch):
    initial_lr = 0.0001
    drop = 0.5
    epochs_drop = 2
    lr = initial_lr * math.pow(drop, math.floor((1+epoch)/epochs_drop))
    print("Learning rate set to:", lr)
    return lr



# Crear un callback para ajustar dinámicamente la tasa de aprendizaje
lr_scheduler = LearningRateScheduler(lr_schedule)

# Mostrar un resumen del modelo que incluye la arquitectura y el número de parámetros
custom_vgg_model.summary()

# Entrenar el modelo utilizando generadores de datos para el conjunto de entrenamiento y validación
model_history = custom_vgg_model.fit(
    train_generator,
    epochs=epochs,
    validation_data=validation_generator,
    steps_per_epoch=nb_train_samples//batch_size,  # Número de pasos por época de entrenamiento
    validation_steps=nb_validation_samples//batch_size,  # Número de pasos por época de validación
    callbacks=[lr_scheduler])# Añadir el callback de la tasa de aprendizaje
   

Model: "functional_3"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_1 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ output (Dense)                  │ (None, 18)             │        73,746 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 134,334,290 (512.44 MB)

 Trainable params: 73,746 (288.07 KB)

 Non-trainable params: 134,260,544 (512.16 MB)

Learning rate set to: 0.0001
Epoch 1/50

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\keras\src\trainers\data_adapters\py_dataset_adapter.py:120: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.
  self._warn_if_super_not_called()

87/87 ━━━━━━━━━━━━━━━━━━━━ 320s 4s/step - accuracy: 0.2627 - loss: 2.9901 - val_accuracy: 0.8281 - val_loss: 0.9816 - learning_rate: 1.0000e-04
Learning rate set to: 5e-05
Epoch 2/50
 1/87 ━━━━━━━━━━━━━━━━━━━━ 5:18 4s/step - accuracy: 0.6875 - loss: 1.3955

C:\Users\Oscar Diaz\anaconda3\Lib\contextlib.py:158: UserWarning: Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches. You may need to use the `.repeat()` function when building your dataset.
  self.gen.throw(typ, value, traceback)

87/87 ━━━━━━━━━━━━━━━━━━━━ 7s 39ms/step - accuracy: 0.6875 - loss: 1.3955 - val_accuracy: 0.6154 - val_loss: 1.3046 - learning_rate: 5.0000e-05
Learning rate set to: 5e-05
Epoch 3/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 314s 4s/step - accuracy: 0.7854 - loss: 1.0984 - val_accuracy: 0.8594 - val_loss: 0.8841 - learning_rate: 5.0000e-05
Learning rate set to: 2.5e-05
Epoch 4/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.7188 - loss: 0.9660 - val_accuracy: 0.9231 - val_loss: 0.7396 - learning_rate: 2.5000e-05
Learning rate set to: 2.5e-05
Epoch 5/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 306s 3s/step - accuracy: 0.8557 - loss: 0.8725 - val_accuracy: 0.8438 - val_loss: 0.8213 - learning_rate: 2.5000e-05
Learning rate set to: 1.25e-05
Epoch 6/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.9062 - loss: 0.6931 - val_accuracy: 0.8846 - val_loss: 0.7597 - learning_rate: 1.2500e-05
Learning rate set to: 1.25e-05
Epoch 7/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 297s 3s/step - accuracy: 0.8782 - loss: 0.8099 - val_accuracy: 0.9219 - val_loss: 0.6788 - learning_rate: 1.2500e-05
Learning rate set to: 6.25e-06
Epoch 8/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.8125 - loss: 0.8384 - val_accuracy: 0.8077 - val_loss: 0.8301 - learning_rate: 6.2500e-06
Learning rate set to: 6.25e-06
Epoch 9/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 297s 3s/step - accuracy: 0.8746 - loss: 0.7692 - val_accuracy: 0.8438 - val_loss: 0.7895 - learning_rate: 6.2500e-06
Learning rate set to: 3.125e-06
Epoch 10/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9375 - loss: 0.5706 - val_accuracy: 0.9231 - val_loss: 0.7769 - learning_rate: 3.1250e-06
Learning rate set to: 3.125e-06
Epoch 11/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 296s 3s/step - accuracy: 0.8766 - loss: 0.7774 - val_accuracy: 0.9062 - val_loss: 0.6919 - learning_rate: 3.1250e-06
Learning rate set to: 1.5625e-06
Epoch 12/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 32ms/step - accuracy: 0.8750 - loss: 0.7578 - val_accuracy: 0.9231 - val_loss: 0.7050 - learning_rate: 1.5625e-06
Learning rate set to: 1.5625e-06
Epoch 13/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 296s 3s/step - accuracy: 0.8920 - loss: 0.7599 - val_accuracy: 0.8594 - val_loss: 0.7321 - learning_rate: 1.5625e-06
Learning rate set to: 7.8125e-07
Epoch 14/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.8750 - loss: 0.8360 - val_accuracy: 0.8462 - val_loss: 0.9521 - learning_rate: 7.8125e-07
Learning rate set to: 7.8125e-07
Epoch 15/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 306s 3s/step - accuracy: 0.8925 - loss: 0.7552 - val_accuracy: 0.9375 - val_loss: 0.6446 - learning_rate: 7.8125e-07
Learning rate set to: 3.90625e-07
Epoch 16/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 29ms/step - accuracy: 0.9062 - loss: 0.6771 - val_accuracy: 0.8846 - val_loss: 0.7051 - learning_rate: 3.9062e-07
Learning rate set to: 3.90625e-07
Epoch 17/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 297s 3s/step - accuracy: 0.8914 - loss: 0.7439 - val_accuracy: 0.9062 - val_loss: 0.7160 - learning_rate: 3.9062e-07
Learning rate set to: 1.953125e-07
Epoch 18/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9062 - loss: 0.7391 - val_accuracy: 0.9231 - val_loss: 0.7109 - learning_rate: 1.9531e-07
Learning rate set to: 1.953125e-07
Epoch 19/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 296s 3s/step - accuracy: 0.8956 - loss: 0.7447 - val_accuracy: 0.9375 - val_loss: 0.7051 - learning_rate: 1.9531e-07
Learning rate set to: 9.765625e-08
Epoch 20/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.8750 - loss: 0.7531 - val_accuracy: 0.7308 - val_loss: 0.7895 - learning_rate: 9.7656e-08
Learning rate set to: 9.765625e-08
Epoch 21/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 302s 3s/step - accuracy: 0.8937 - loss: 0.7510 - val_accuracy: 0.9375 - val_loss: 0.6711 - learning_rate: 9.7656e-08
Learning rate set to: 4.8828125e-08
Epoch 22/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.9688 - loss: 0.6780 - val_accuracy: 0.9615 - val_loss: 0.6467 - learning_rate: 4.8828e-08
Learning rate set to: 4.8828125e-08
Epoch 23/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 297s 3s/step - accuracy: 0.8882 - loss: 0.7567 - val_accuracy: 0.8750 - val_loss: 0.6856 - learning_rate: 4.8828e-08
Learning rate set to: 2.44140625e-08
Epoch 24/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.7500 - loss: 0.9866 - val_accuracy: 0.8462 - val_loss: 0.8062 - learning_rate: 2.4414e-08
Learning rate set to: 2.44140625e-08
Epoch 25/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.8869 - loss: 0.7534 - val_accuracy: 0.9219 - val_loss: 0.6431 - learning_rate: 2.4414e-08
Learning rate set to: 1.220703125e-08
Epoch 26/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9375 - loss: 0.6891 - val_accuracy: 0.8462 - val_loss: 0.7317 - learning_rate: 1.2207e-08
Learning rate set to: 1.220703125e-08
Epoch 27/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 296s 3s/step - accuracy: 0.8814 - loss: 0.7565 - val_accuracy: 0.9062 - val_loss: 0.7840 - learning_rate: 1.2207e-08
Learning rate set to: 6.103515625e-09
Epoch 28/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 7s 32ms/step - accuracy: 0.9375 - loss: 0.6982 - val_accuracy: 0.8846 - val_loss: 0.6782 - learning_rate: 6.1035e-09
Learning rate set to: 6.103515625e-09
Epoch 29/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.8971 - loss: 0.7462 - val_accuracy: 0.9375 - val_loss: 0.6778 - learning_rate: 6.1035e-09
Learning rate set to: 3.0517578125e-09
Epoch 30/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 7s 40ms/step - accuracy: 0.8750 - loss: 0.8374 - val_accuracy: 0.9615 - val_loss: 0.6405 - learning_rate: 3.0518e-09
Learning rate set to: 3.0517578125e-09
Epoch 31/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 298s 3s/step - accuracy: 0.9021 - loss: 0.7392 - val_accuracy: 0.9219 - val_loss: 0.7032 - learning_rate: 3.0518e-09
Learning rate set to: 1.52587890625e-09
Epoch 32/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.7812 - loss: 0.7991 - val_accuracy: 0.9615 - val_loss: 0.6517 - learning_rate: 1.5259e-09
Learning rate set to: 1.52587890625e-09
Epoch 33/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 297s 3s/step - accuracy: 0.8967 - loss: 0.7477 - val_accuracy: 0.8906 - val_loss: 0.7329 - learning_rate: 1.5259e-09
Learning rate set to: 7.62939453125e-10
Epoch 34/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 7s 38ms/step - accuracy: 0.9688 - loss: 0.6851 - val_accuracy: 0.8846 - val_loss: 0.8029 - learning_rate: 7.6294e-10
Learning rate set to: 7.62939453125e-10
Epoch 35/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 301s 3s/step - accuracy: 0.8865 - loss: 0.7543 - val_accuracy: 0.8750 - val_loss: 0.7026 - learning_rate: 7.6294e-10
Learning rate set to: 3.814697265625e-10
Epoch 36/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.8438 - loss: 0.7618 - val_accuracy: 1.0000 - val_loss: 0.5661 - learning_rate: 3.8147e-10
Learning rate set to: 3.814697265625e-10
Epoch 37/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 297s 3s/step - accuracy: 0.8912 - loss: 0.7499 - val_accuracy: 0.9219 - val_loss: 0.7116 - learning_rate: 3.8147e-10
Learning rate set to: 1.9073486328125e-10
Epoch 38/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 32ms/step - accuracy: 0.8438 - loss: 0.7884 - val_accuracy: 0.9615 - val_loss: 0.6267 - learning_rate: 1.9073e-10
Learning rate set to: 1.9073486328125e-10
Epoch 39/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 300s 3s/step - accuracy: 0.8936 - loss: 0.7434 - val_accuracy: 0.8594 - val_loss: 0.6922 - learning_rate: 1.9073e-10
Learning rate set to: 9.5367431640625e-11
Epoch 40/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.8750 - loss: 0.7543 - val_accuracy: 0.9231 - val_loss: 0.7359 - learning_rate: 9.5367e-11
Learning rate set to: 9.5367431640625e-11
Epoch 41/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 297s 3s/step - accuracy: 0.9073 - loss: 0.7348 - val_accuracy: 0.8594 - val_loss: 0.7475 - learning_rate: 9.5367e-11
Learning rate set to: 4.76837158203125e-11
Epoch 42/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9062 - loss: 0.7780 - val_accuracy: 0.9231 - val_loss: 0.6189 - learning_rate: 4.7684e-11
Learning rate set to: 4.76837158203125e-11
Epoch 43/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 296s 3s/step - accuracy: 0.8979 - loss: 0.7500 - val_accuracy: 0.9531 - val_loss: 0.6084 - learning_rate: 4.7684e-11
Learning rate set to: 2.384185791015625e-11
Epoch 44/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.9062 - loss: 0.7015 - val_accuracy: 0.8077 - val_loss: 0.9279 - learning_rate: 2.3842e-11
Learning rate set to: 2.384185791015625e-11
Epoch 45/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 296s 3s/step - accuracy: 0.9023 - loss: 0.7475 - val_accuracy: 0.9062 - val_loss: 0.7125 - learning_rate: 2.3842e-11
Learning rate set to: 1.1920928955078126e-11
Epoch 46/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.8125 - loss: 0.9087 - val_accuracy: 0.8846 - val_loss: 0.8064 - learning_rate: 1.1921e-11
Learning rate set to: 1.1920928955078126e-11
Epoch 47/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.8690 - loss: 0.8052 - val_accuracy: 0.8750 - val_loss: 0.7635 - learning_rate: 1.1921e-11
Learning rate set to: 5.960464477539063e-12
Epoch 48/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9688 - loss: 0.7053 - val_accuracy: 0.9615 - val_loss: 0.5546 - learning_rate: 5.9605e-12
Learning rate set to: 5.960464477539063e-12
Epoch 49/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 302s 3s/step - accuracy: 0.8988 - loss: 0.7302 - val_accuracy: 0.9531 - val_loss: 0.6477 - learning_rate: 5.9605e-12
Learning rate set to: 2.9802322387695314e-12
Epoch 50/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.9062 - loss: 0.9430 - val_accuracy: 0.9615 - val_loss: 0.6262 - learning_rate: 2.9802e-12

Este código realiza las siguientes operaciones en general:

1. Definición de variables: Se definen dos variables nb_train_samples y nb_validation_samples para indicar el número de muestras de entrenamiento y validación respectivamente. Estas variables se utilizan más adelante para configurar el entrenamiento del modelo.

2. Definición de la entrada de la red neuronal: Se define la entrada de la red neuronal utilizando la clase Input de Keras. La forma de la entrada se especifica como (width_shape, height_shape, 3), lo que indica el tamaño de las imágenes de entrada y el número de canales de color (en este caso, 3 para imágenes RGB).

3. Carga del modelo preentrenado VGG16: Se carga el modelo preentrenado VGG16 utilizando la función VGG16 de Keras. Se especifica el tensor de entrada de la red neuronal (input_tensor=image_input), se incluye la capa densa al final del modelo (include_top=True) y se cargan los pesos preentrenados en el conjunto de datos de ImageNet (weights='imagenet').

4. Creación de un nuevo modelo personalizado: Se crea un nuevo modelo personalizado que incluye todas las capas de la VGG16 hasta la última capa densa (fc2). Se agrega una capa densa adicional al final del modelo con un número de unidades igual al número de clases en el problema (num_classes). Esta capa tiene una función de activación softmax, lo que la convierte en una clasificación de salida.

5. Congelación de capas: Se congela todas las capas del modelo preentrenado VGG16, excepto la última capa densa añadida. Esto se hace para que durante el entrenamiento solo se actualicen los pesos de la nueva capa densa.

6. Compilación del modelo: Se compila el modelo personalizado utilizando la función de pérdida categorical_crossentropy, el optimizador adadelta y se incluye la métrica de precisión (accuracy).

7. Resumen del modelo: Se muestra un resumen del modelo, que incluye todas las capas y el número total de parámetros entrenables y no entrenables.

8. Entrenamiento del modelo: Se entrena el modelo utilizando los generadores de datos train_generator y validation_generator que se definieron anteriormente. Se especifica el número de épocas (epochs) y el número de pasos por época (steps_per_epoch) y de validación (validation_steps). Durante el entrenamiento, se actualizan los pesos de la capa densa añadida mientras que los pesos de las capas de VGG16 se mantienen fijos debido a la congelación realizada anteriormente.

Grabar modelo en disco

Guardar el modelo tiene varias ventajas. En primer lugar, permite la reutilización futura, ya que puedes cargarlo y utilizarlo nuevamente sin necesidad de volver a entrenarlo desde cero, lo que resulta útil tanto para hacer predicciones en nuevos datos como para continuar el entrenamiento en una fecha posterior. Además, facilita la distribución y compartición del modelo con otros investigadores, colegas o clientes que puedan necesitar utilizarlo en sus propios proyectos. Por último, al guardar el modelo junto con su configuración y pesos entrenados, se asegura la reproducibilidad de los resultados, ya que otros investigadores pueden cargar el modelo exacto y obtener los mismos resultados que tú, lo que es fundamental para la validación y la comparación de resultados en investigación científica y desarrollo de modelos.

Este código utilizará un bucle while para verificar si el archivo con el nombre base del modelo ya existe. Si existe, agregará un número al final del nombre del archivo y verificará nuevamente. Esto continuará hasta que se encuentre un nombre de archivo único que no exista en el directorio. Una vez que se encuentra un nombre único, el modelo se guarda con ese nombre.

import os

# Nombre base del modelo
model_name = "model_VGG16_v"

# Extensión del archivo
file_extension = ".keras"

# Directorio donde se guardarán los modelos
model_directory = "models/"

# Inicializar contador
counter = 1

# Generar el nombre completo del archivo
file_name = model_name + file_extension

ruta=model_directory + file_name
print(ruta)
# Verificar si el modelo ya está guardado
while os.path.exists(model_directory + file_name):
    
    # Si el archivo existe, agregar un número al final del nombre del modelo
    file_name = f"{model_name}{counter}{file_extension}"
    counter += 1

# Guardar el modelo con el nombre único en el directorio correcto
custom_vgg_model.save(model_directory + file_name)

models/model_VGG16_v.keras

Gráficas de entrenamiento y validación (accuracy - loss)

def plotTraining(hist, epochs, typeData):
    
    if typeData=="loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'-r',label='Loss Training')
    if typeData=="accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'-r',label='Accuracy Training')
    if typeData=="val_loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['val_loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'--b',label='Loss Validate')
    if typeData=="val_accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['val_accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'--b',label='Training Validate')
        

    plt.rc('xtick',labelsize=24)
    plt.rc('ytick',labelsize=24)
    plt.rc('legend', fontsize=18) 
    plt.legend()
    plt.xlabel('Number of Epochs',fontsize=24)
    plt.grid(True)

plotTraining(model_history,epochs,"loss")
plotTraining(model_history,epochs,"accuracy")
plotTraining(model_history,epochs,"val_loss")
plotTraining(model_history,epochs,"val_accuracy")

Predicción usando el modelo entrenado

from keras.applications.imagenet_utils import preprocess_input, decode_predictions
from keras.models import load_model

names = ['CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS', 'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS', 'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA', 'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']

# Cargar el modelo
modelt = load_model(model_directory + file_name)
#modelt = custom_vgg_model

# Ruta de la imagen de prueba
imaget_path = "ImagenPrueba_SF.png"

# Leer la imagen, cambiar tamaño y preprocesar
imaget=cv2.resize(cv2.imread(imaget_path), (width_shape, height_shape), interpolation = cv2.INTER_AREA)
xt = np.asarray(imaget)
xt=preprocess_input(xt)
xt = np.expand_dims(xt,axis=0)

# Obtener las predicciones del modelo
preds = modelt.predict(xt)

# Obtener la clase predicha y su porcentaje de confianza
predicted_class_index = np.argmax(preds)
predicted_class_name = names[predicted_class_index]
confidence_percentage = preds[0][predicted_class_index] * 100

# Imprimir el resultado
print(f'Clase predicha: {predicted_class_name}')
print(f'Porcentaje de confianza: {confidence_percentage:.2f}%')

# Mostrar la imagen
plt.imshow(cv2.cvtColor(np.asarray(imaget), cv2.COLOR_BGR2RGB))
plt.axis('off')
plt.show()

1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 656ms/step
Clase predicha: CYANOCORAX YNCAS
Porcentaje de confianza: 89.46%

Matriz de confusión y métricas de desempeño

from sklearn.metrics import confusion_matrix, f1_score, roc_curve, precision_score, recall_score, accuracy_score, roc_auc_score
from sklearn import metrics
from mlxtend.plotting import plot_confusion_matrix
from keras.models import load_model
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline


names = ['CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS', 'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS', 'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA', 'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']



test_data_dir = 'dataset/test'  

test_datagen = ImageDataGenerator()

test_generator = test_datagen.flow_from_directory(
    test_data_dir,
    target_size=(width_shape, height_shape), 
    batch_size = batch_size,
    class_mode='categorical', 
    shuffle=False)

custom_Model= load_model(model_directory + file_name)

predictions = custom_Model.predict(test_generator)


y_pred = np.argmax(predictions, axis=1)
y_real = test_generator.classes


matc=confusion_matrix(y_real, y_pred)

plot_confusion_matrix(conf_mat=matc, figsize=(9,9), class_names = names, show_normed=False)
plt.tight_layout()

print(metrics.classification_report(y_real,y_pred, digits = 4))

Found 86 images belonging to 18 classes.
3/3 ━━━━━━━━━━━━━━━━━━━━ 9s 3s/step

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\mlxtend\plotting\plot_confusion_matrix.py:102: RuntimeWarning: invalid value encountered in divide
  normed_conf_mat = conf_mat.astype("float") / total_samples
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))

              precision    recall  f1-score   support

           0     1.0000    0.2000    0.3333         5
           1     0.5714    0.8000    0.6667         5
           2     0.8333    1.0000    0.9091         5
           3     1.0000    1.0000    1.0000         5
           4     0.6250    1.0000    0.7692         5
           5     0.7500    1.0000    0.8571         6
           6     0.4444    0.8000    0.5714         5
           7     0.4167    1.0000    0.5882         5
           8     1.0000    0.8000    0.8889         5
           9     0.1250    0.2000    0.1538         5
          10     0.5000    0.2000    0.2857         5
          11     0.2500    0.2000    0.2222         5
          12     0.0000    0.0000    0.0000         5
          13     1.0000    0.4000    0.5714         5
          14     1.0000    0.6000    0.7500         5
          15     1.0000    0.2000    0.3333         5
          16     0.0000    0.0000    0.0000         5
          17     0.0000    0.0000    0.0000         0

    accuracy                         0.5581        86
   macro avg     0.5842    0.5222    0.4945        86
weighted avg     0.6201    0.5581    0.5274        86

# Define el tamaño de la figura y los nombres de las clases
fig, ax = plot_confusion_matrix(conf_mat=matc, figsize=(9,9), class_names=names, show_normed=True)

# Ajusta el tamaño de las letras de los nombres de las clases
ax.set_xticklabels(names, fontsize=12)  # Ajusta el tamaño de las letras en el eje x
ax.set_yticklabels(names, fontsize=12)  # Ajusta el tamaño de las letras en el eje y

# Ajusta automáticamente el diseño de la figura
plt.tight_layout()

# Muestra la figura
plt.show()

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\mlxtend\plotting\plot_confusion_matrix.py:102: RuntimeWarning: invalid value encountered in divide
  normed_conf_mat = conf_mat.astype("float") / total_samples

Prueba 3

Regularización L2

Entrenamiento

# Importar las funciones necesarias de Keras
from keras.layers import Input
from keras.applications import VGG16

# Definir la forma de entrada para las imágenes (ancho, alto, canales RGB)
image_input = Input(shape=(width_shape, height_shape, 3))

# Crear el modelo VGG16 utilizando la entrada de imagen definida
# include_top=True significa que se incluirán todas las capas densas en la parte superior del modelo
# weights='imagenet' significa que se utilizarán los pesos pre-entrenados en ImageNet
model2 = VGG16(input_tensor=image_input, include_top=True, weights='imagenet')

# Mostrar un resumen de la arquitectura del modelo
model2.summary()

Model: "vgg16"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_2 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,357,544 (527.79 MB)

 Trainable params: 138,357,544 (527.79 MB)

 Non-trainable params: 0 (0.00 B)

last_layer = model2.get_layer('block5_pool').output
x= Flatten(name='flatten')(last_layer)
x = Dense(128, activation='relu', name='fc1')(x)
x = Dense(128, activation='relu', name='fc2')(x)
out = Dense(num_classes, activation='softmax', name='output')(x)
custom_model = Model(image_input, out)
custom_model.summary()

# freeze all the layers except the dense layers
for layer in custom_model.layers[:-3]:
	layer.trainable = False

custom_model.summary()

custom_model.compile(loss='categorical_crossentropy',optimizer='adadelta',metrics=['accuracy'])

Model: "functional_5"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_2 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 128)            │     3,211,392 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 128)            │        16,512 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ output (Dense)                  │ (None, 18)             │         2,322 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 17,944,914 (68.45 MB)

 Trainable params: 17,944,914 (68.45 MB)

 Non-trainable params: 0 (0.00 B)

Model: "functional_5"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_2 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 128)            │     3,211,392 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 128)            │        16,512 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ output (Dense)                  │ (None, 18)             │         2,322 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 17,944,914 (68.45 MB)

 Trainable params: 3,230,226 (12.32 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

# Definir el número de muestras de entrenamiento y validación
nb_train_samples = 2805
nb_validation_samples = 90

model_history = custom_model.fit(  
    train_generator,
    epochs=epochs,
    validation_data=validation_generator,
    steps_per_epoch=nb_train_samples//batch_size,
    validation_steps=nb_validation_samples//batch_size)

Epoch 1/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 298s 3s/step - accuracy: 0.0663 - loss: 18.7191 - val_accuracy: 0.0312 - val_loss: 18.9593
Epoch 2/50
 1/87 ━━━━━━━━━━━━━━━━━━━━ 4:15 3s/step - accuracy: 0.0625 - loss: 17.2532

C:\Users\Oscar Diaz\anaconda3\Lib\contextlib.py:158: UserWarning: Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches. You may need to use the `.repeat()` function when building your dataset.
  self.gen.throw(typ, value, traceback)

87/87 ━━━━━━━━━━━━━━━━━━━━ 5s 29ms/step - accuracy: 0.0625 - loss: 17.2532 - val_accuracy: 0.1154 - val_loss: 12.6727
Epoch 3/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 293s 3s/step - accuracy: 0.0679 - loss: 15.7419 - val_accuracy: 0.1562 - val_loss: 13.0466
Epoch 4/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 32ms/step - accuracy: 0.1250 - loss: 16.9887 - val_accuracy: 0.0769 - val_loss: 14.3917
Epoch 5/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 307s 4s/step - accuracy: 0.0892 - loss: 13.6572 - val_accuracy: 0.0938 - val_loss: 14.8485
Epoch 6/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.0312 - loss: 12.3036 - val_accuracy: 0.1538 - val_loss: 9.1232
Epoch 7/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 296s 3s/step - accuracy: 0.1002 - loss: 12.6286 - val_accuracy: 0.1406 - val_loss: 11.9384
Epoch 8/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.0625 - loss: 12.5009 - val_accuracy: 0.1538 - val_loss: 10.8247
Epoch 9/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 301s 3s/step - accuracy: 0.1010 - loss: 11.3096 - val_accuracy: 0.0938 - val_loss: 11.0147
Epoch 10/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 29ms/step - accuracy: 0.1250 - loss: 8.9476 - val_accuracy: 0.0769 - val_loss: 11.6915
Epoch 11/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 293s 3s/step - accuracy: 0.1191 - loss: 10.0668 - val_accuracy: 0.1719 - val_loss: 10.0200
Epoch 12/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.0625 - loss: 11.1884 - val_accuracy: 0.1154 - val_loss: 10.3086
Epoch 13/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 294s 3s/step - accuracy: 0.1092 - loss: 9.7454 - val_accuracy: 0.1250 - val_loss: 9.8759
Epoch 14/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 29ms/step - accuracy: 0.1250 - loss: 9.2793 - val_accuracy: 0.0000e+00 - val_loss: 12.3645
Epoch 15/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 293s 3s/step - accuracy: 0.1582 - loss: 9.0621 - val_accuracy: 0.1875 - val_loss: 8.1522
Epoch 16/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 29ms/step - accuracy: 0.1875 - loss: 9.3314 - val_accuracy: 0.0769 - val_loss: 11.2243
Epoch 17/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 293s 3s/step - accuracy: 0.1632 - loss: 8.6203 - val_accuracy: 0.2188 - val_loss: 7.7420
Epoch 18/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.0625 - loss: 8.3507 - val_accuracy: 0.1538 - val_loss: 6.6532
Epoch 19/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 293s 3s/step - accuracy: 0.1686 - loss: 8.0066 - val_accuracy: 0.0938 - val_loss: 9.3384
Epoch 20/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 29ms/step - accuracy: 0.1562 - loss: 6.8174 - val_accuracy: 0.3077 - val_loss: 7.0244
Epoch 21/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 294s 3s/step - accuracy: 0.1580 - loss: 7.6412 - val_accuracy: 0.1250 - val_loss: 8.8477
Epoch 22/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.2812 - loss: 6.1855 - val_accuracy: 0.1923 - val_loss: 7.5021
Epoch 23/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.2020 - loss: 6.8421 - val_accuracy: 0.2031 - val_loss: 7.6823
Epoch 24/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.2812 - loss: 6.8964 - val_accuracy: 0.1154 - val_loss: 7.3148
Epoch 25/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 300s 3s/step - accuracy: 0.2051 - loss: 6.7834 - val_accuracy: 0.1406 - val_loss: 7.8871
Epoch 26/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 29ms/step - accuracy: 0.2500 - loss: 6.7696 - val_accuracy: 0.1538 - val_loss: 7.4677
Epoch 27/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.1910 - loss: 6.4696 - val_accuracy: 0.3125 - val_loss: 5.7271
Epoch 28/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.2500 - loss: 6.1857 - val_accuracy: 0.1538 - val_loss: 7.6506
Epoch 29/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 294s 3s/step - accuracy: 0.1982 - loss: 6.3297 - val_accuracy: 0.1719 - val_loss: 6.8591
Epoch 30/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.2500 - loss: 5.6456 - val_accuracy: 0.3077 - val_loss: 4.8336
Epoch 31/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 297s 3s/step - accuracy: 0.2154 - loss: 5.9391 - val_accuracy: 0.1875 - val_loss: 6.4135
Epoch 32/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.2188 - loss: 5.3871 - val_accuracy: 0.1538 - val_loss: 5.8893
Epoch 33/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 293s 3s/step - accuracy: 0.2254 - loss: 5.7190 - val_accuracy: 0.2344 - val_loss: 5.3781
Epoch 34/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.1875 - loss: 6.2125 - val_accuracy: 0.3077 - val_loss: 5.1460
Epoch 35/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 294s 3s/step - accuracy: 0.2335 - loss: 5.5416 - val_accuracy: 0.1875 - val_loss: 5.7378
Epoch 36/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 32ms/step - accuracy: 0.1875 - loss: 5.5248 - val_accuracy: 0.3846 - val_loss: 7.0313
Epoch 37/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.2353 - loss: 5.3585 - val_accuracy: 0.1875 - val_loss: 7.0348
Epoch 38/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.3125 - loss: 5.2204 - val_accuracy: 0.1538 - val_loss: 6.5614
Epoch 39/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 294s 3s/step - accuracy: 0.2596 - loss: 5.1959 - val_accuracy: 0.2656 - val_loss: 5.9645
Epoch 40/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.3750 - loss: 3.3942 - val_accuracy: 0.1923 - val_loss: 4.2173
Epoch 41/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 301s 3s/step - accuracy: 0.2552 - loss: 4.8990 - val_accuracy: 0.2656 - val_loss: 4.7025
Epoch 42/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.2188 - loss: 5.3079 - val_accuracy: 0.3462 - val_loss: 4.1265
Epoch 43/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.2820 - loss: 4.8741 - val_accuracy: 0.2812 - val_loss: 4.8116
Epoch 44/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.2812 - loss: 5.4825 - val_accuracy: 0.2308 - val_loss: 4.4658
Epoch 45/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 295s 3s/step - accuracy: 0.2944 - loss: 4.4461 - val_accuracy: 0.3438 - val_loss: 4.4990
Epoch 46/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.2188 - loss: 4.8492 - val_accuracy: 0.3462 - val_loss: 4.6156
Epoch 47/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 296s 3s/step - accuracy: 0.2920 - loss: 4.5979 - val_accuracy: 0.2969 - val_loss: 5.1433
Epoch 48/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.6250 - loss: 2.6970 - val_accuracy: 0.2308 - val_loss: 5.2890
Epoch 49/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 293s 3s/step - accuracy: 0.3138 - loss: 4.2803 - val_accuracy: 0.2500 - val_loss: 4.7989
Epoch 50/50
87/87 ━━━━━━━━━━━━━━━━━━━━ 6s 30ms/step - accuracy: 0.2812 - loss: 5.2754 - val_accuracy: 0.4231 - val_loss: 4.1803

Grabar modelo en disco

import os

# Nombre base del modelo
model_name = "model_VGG16_v"

# Extensión del archivo
file_extension = ".keras"

# Directorio donde se guardarán los modelos
model_directory = "models/"

# Inicializar contador
counter = 1

# Generar el nombre completo del archivo
file_name = model_name + file_extension

ruta=model_directory + file_name
print(ruta)
# Verificar si el modelo ya está guardado
while os.path.exists(model_directory + file_name):
    
    # Si el archivo existe, agregar un número al final del nombre del modelo
    file_name = f"{model_name}{counter}{file_extension}"
    counter += 1

# Guardar el modelo con el nombre único en el directorio correcto
custom_vgg_model.save(model_directory + file_name)

models/model_VGG16_v.keras

Gráficas de entrenamiento y validación (accuracy - loss)

def plotTraining(hist, epochs, typeData):
    
    if typeData=="loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'-r',label='Loss Training')
    if typeData=="accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'-r',label='Accuracy Training')
    if typeData=="val_loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['val_loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'--b',label='Loss Validate')
    if typeData=="val_accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['val_accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'--b',label='Training Validate')
        

    plt.rc('xtick',labelsize=24)
    plt.rc('ytick',labelsize=24)
    plt.rc('legend', fontsize=18) 
    plt.legend()
    plt.xlabel('Number of Epochs',fontsize=24)
    plt.grid(True)

plotTraining(model_history,epochs,"loss")
plotTraining(model_history,epochs,"accuracy")
plotTraining(model_history,epochs,"val_loss")
plotTraining(model_history,epochs,"val_accuracy")

Predicción usando el modelo entrenado

from keras.applications.imagenet_utils import preprocess_input, decode_predictions
from keras.models import load_model

names = ['CYANOCORAX YNCAS', 'PIRANGA RUBRA', 'PITANGUS SULPHURATUS', 'PYROCEPHALUS RUBINUS', 'RUPORNIS MAGNIROSTRIS', 'SICALIS FLAVEOLA', 'THRAUPIS EPISCOPUS', 'TIARIS OLIVACEUS', 'TYRANNUS MELANCHOLICUS', 'ZONOTRICHIA CAPENSIS', 'CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS', 'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS', 'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA', 'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']

# Cargar el modelo
modelt = load_model(model_directory + file_name)
#modelt = custom_vgg_model

# Ruta de la imagen de prueba
imaget_path = "ImagenPrueba_SF.png"

# Leer la imagen, cambiar tamaño y preprocesar
imaget=cv2.resize(cv2.imread(imaget_path), (width_shape, height_shape), interpolation = cv2.INTER_AREA)
xt = np.asarray(imaget)
xt=preprocess_input(xt)
xt = np.expand_dims(xt,axis=0)

# Obtener las predicciones del modelo
preds = modelt.predict(xt)

# Obtener la clase predicha y su porcentaje de confianza
predicted_class_index = np.argmax(preds)
predicted_class_name = names[predicted_class_index]
confidence_percentage = preds[0][predicted_class_index] * 100

# Imprimir el resultado
print(f'Clase predicha: {predicted_class_name}')
print(f'Porcentaje de confianza: {confidence_percentage:.2f}%')

# Mostrar la imagen
plt.imshow(cv2.cvtColor(np.asarray(imaget), cv2.COLOR_BGR2RGB))
plt.axis('off')
plt.show()

1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 743ms/step
Clase predicha: SICALIS FLAVEOLA
Porcentaje de confianza: 89.46%

Matriz de confusión y métricas de desempeño

La validación oscila tanto porque podría haber un problema de sobreajuste (overfitting) o la necesidad de ajustar los hiperparámetros. El modelo parece estar aprendiendo bien durante el entrenamiento, como se muestra en la línea roja continua “Accuracy Training”, pero tiene un rendimiento inconsistente en los datos de validación, indicado por la línea azul punteada “Training Validate”.

Para abordar esto:

Regularización: Introduce técnicas de regularización como dropout o weight decay para reducir el sobreajuste. Ajuste de Hiperparámetros: Experimenta con diferentes valores para hiperparámetros como tasa de aprendizaje, número de capas ocultas, etc. Más Datos: Si es posible, aumenta el tamaño del conjunto de datos de entrenamiento. Early Stopping: Detén el entrenamiento cuando la precisión de validación deja de mejorar. El gráfico muestra que el modelo necesita más ajustes para generalizar mejor a datos no vistos

from sklearn.metrics import confusion_matrix, f1_score, roc_curve, precision_score, recall_score, accuracy_score, roc_auc_score
from sklearn import metrics
from mlxtend.plotting import plot_confusion_matrix
from keras.models import load_model
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline


names = [ 'CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS', 'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS', 'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA', 'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']



test_data_dir = 'dataset/test'  

test_datagen = ImageDataGenerator()

test_generator = test_datagen.flow_from_directory(
    test_data_dir,
    target_size=(width_shape, height_shape), 
    batch_size = batch_size,
    class_mode='categorical', 
    shuffle=False)

custom_Model50= load_model(model_directory + file_name)

predictions = custom_Model50.predict(test_generator)


y_pred = np.argmax(predictions, axis=1)
y_real = test_generator.classes


matc=confusion_matrix(y_real, y_pred)

plot_confusion_matrix(conf_mat=matc, figsize=(9,9), class_names = names, show_normed=False)
plt.tight_layout()

print(metrics.classification_report(y_real,y_pred, digits = 4))

Found 89 images belonging to 18 classes.

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\keras\src\trainers\data_adapters\py_dataset_adapter.py:120: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.
  self._warn_if_super_not_called()

3/3 ━━━━━━━━━━━━━━━━━━━━ 13s 4s/step

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\mlxtend\plotting\plot_confusion_matrix.py:102: RuntimeWarning: invalid value encountered in divide
  normed_conf_mat = conf_mat.astype("float") / total_samples
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Precision and F-score are ill-defined and being set to 0.0 in labels with no predicted samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))
C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\sklearn\metrics\_classification.py:1344: UndefinedMetricWarning: Recall and F-score are ill-defined and being set to 0.0 in labels with no true samples. Use `zero_division` parameter to control this behavior.
  _warn_prf(average, modifier, msg_start, len(result))

              precision    recall  f1-score   support

           0     1.0000    0.2000    0.3333         5
           1     0.5714    0.8000    0.6667         5
           2     0.8333    1.0000    0.9091         5
           3     1.0000    1.0000    1.0000         5
           4     0.6250    1.0000    0.7692         5
           5     0.7500    1.0000    0.8571         6
           6     0.5833    0.8750    0.7000         8
           7     0.4167    1.0000    0.5882         5
           8     1.0000    0.8000    0.8889         5
           9     0.1250    0.2000    0.1538         5
          10     0.5000    0.2000    0.2857         5
          11     0.2500    0.2000    0.2222         5
          12     0.0000    0.0000    0.0000         5
          13     1.0000    0.4000    0.5714         5
          14     1.0000    0.6000    0.7500         5
          15     1.0000    0.2000    0.3333         5
          16     0.0000    0.0000    0.0000         5
          17     0.0000    0.0000    0.0000         0

    accuracy                         0.5730        89
   macro avg     0.5919    0.5264    0.5016        89
weighted avg     0.6267    0.5730    0.5405        89

Prueba 4

width_shape = 224
height_shape = 224
num_classes = 5
epochs = 50
batch_size = 32

train_data_dir = 'dataset/train'  
validation_data_dir = 'dataset/valid'

# Definir el generador de imágenes para el conjunto de entrenamiento con aumentos de datos
train_datagen = ImageDataGenerator(  
    rotation_range=20,               # Rango de grados para rotación aleatoria
    zoom_range=0.2,                  # Rango de zoom aleatorio
    width_shift_range=0.1,           # Rango de desplazamiento horizontal aleatorio
    height_shift_range=0.1,          # Rango de desplazamiento vertical aleatorio
    horizontal_flip=True,            # Volteo horizontal aleatorio
    vertical_flip=False,             # No se aplica volteo vertical
    preprocessing_function=preprocess_input)  # Función de preprocesamiento

# Definir el generador de imágenes para el conjunto de validación con los mismos aumentos de datos
valid_datagen = ImageDataGenerator(    
    rotation_range=20,
    zoom_range=0.2,
    width_shift_range=0.1,
    height_shift_range=0.1,
    horizontal_flip=True,
    vertical_flip=False,
    preprocessing_function=preprocess_input)

# Crear un generador de lotes de imágenes para el conjunto de entrenamiento
train_generator = train_datagen.flow_from_directory(  
    train_data_dir,                  # Directorio que contiene las imágenes de entrenamiento
    target_size=(width_shape, height_shape),  # Tamaño al que se redimensionarán las imágenes
    batch_size=batch_size,           # Tamaño del lote
    class_mode='categorical')        # Modo de clasificación para imágenes categóricas

# Crear un generador de lotes de imágenes para el conjunto de validación
validation_generator = valid_datagen.flow_from_directory(  
    validation_data_dir,             # Directorio que contiene las imágenes de validación
    target_size=(width_shape, height_shape),  # Tamaño al que se redimensionarán las imágenes
    batch_size=batch_size,           # Tamaño del lote
    class_mode='categorical')        # Modo de clasificación para imágenes categóricas

Found 834 images belonging to 5 classes.
Found 25 images belonging to 5 classes.

# Definir el número de muestras de entrenamiento y validación
nb_train_samples = 834
nb_validation_samples = 25

# Definir la entrada de la red neuronal con el tamaño de las imágenes
image_input = Input(shape=(width_shape, height_shape, 3))

# Cargar el modelo VGG16 preentrenado con pesos ajustados desde ImageNet
model = VGG16(input_tensor=image_input, include_top=True, weights='imagenet')

# Obtener la salida de la penúltima capa densa del modelo VGG16 (fc2)
last_layer = model.get_layer('fc2').output

# Añadir una nueva capa densa al final del modelo para la clasificación multiclase con regularización L2 (Evita sobreajuste)
out = Dense(num_classes, activation='softmax', kernel_regularizer='l2', name='output')(last_layer)

# Crear un nuevo modelo personalizado que toma la entrada de la imagen y produce la salida clasificada
custom_vgg_model = Model(image_input, out)

# Congelar todas las capas del modelo, excepto la capa densa añadida
for layer in custom_vgg_model.layers[:-1]:
    layer.trainable = False

# Compilar el modelo con una función de pérdida, optimizador y métricas especificadas
custom_vgg_model.compile(loss='categorical_crossentropy', optimizer=Adam(learning_rate=0.0001), metrics=['accuracy'])


# Mostrar un resumen del modelo que incluye la arquitectura y el número de parámetros
custom_vgg_model.summary()

# Entrenar el modelo utilizando generadores de datos para el conjunto de entrenamiento y validación
model_history = custom_vgg_model.fit(
    train_generator,
    epochs=epochs,
    validation_data=validation_generator,
    steps_per_epoch=nb_train_samples//batch_size,  # Número de pasos por época de entrenamiento
    validation_steps=nb_validation_samples//batch_size)  # Número de pasos por época de validación

Model: "functional_7"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_3 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ output (Dense)                  │ (None, 5)              │        20,485 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 134,281,029 (512.24 MB)

 Trainable params: 20,485 (80.02 KB)

 Non-trainable params: 134,260,544 (512.16 MB)

Epoch 1/50

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\keras\src\trainers\data_adapters\py_dataset_adapter.py:120: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.
  self._warn_if_super_not_called()

26/26 ━━━━━━━━━━━━━━━━━━━━ 91s 3s/step - accuracy: 0.2838 - loss: 2.2863 - val_accuracy: 0.6000 - val_loss: 1.0493
Epoch 2/50
 1/26 ━━━━━━━━━━━━━━━━━━━━ 1:17 3s/step - accuracy: 0.6562 - loss: 1.1432

C:\Users\Oscar Diaz\anaconda3\Lib\contextlib.py:158: UserWarning: Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches. You may need to use the `.repeat()` function when building your dataset.
  self.gen.throw(typ, value, traceback)

26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 114ms/step - accuracy: 0.6562 - loss: 1.1432 - val_accuracy: 0.6000 - val_loss: 1.0793
Epoch 3/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 89s 3s/step - accuracy: 0.7356 - loss: 0.8568 - val_accuracy: 0.8400 - val_loss: 0.5498
Epoch 4/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 114ms/step - accuracy: 0.7812 - loss: 0.5293 - val_accuracy: 0.7600 - val_loss: 0.7074
Epoch 5/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 87s 3s/step - accuracy: 0.8678 - loss: 0.5005 - val_accuracy: 0.8800 - val_loss: 0.4185
Epoch 6/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 121ms/step - accuracy: 0.9375 - loss: 0.2667 - val_accuracy: 0.8000 - val_loss: 0.5027
Epoch 7/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 88s 3s/step - accuracy: 0.9241 - loss: 0.3500 - val_accuracy: 0.9200 - val_loss: 0.3046
Epoch 8/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 123ms/step - accuracy: 0.9375 - loss: 0.3103 - val_accuracy: 0.8400 - val_loss: 0.4757
Epoch 9/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 94s 4s/step - accuracy: 0.9409 - loss: 0.3014 - val_accuracy: 0.8800 - val_loss: 0.2997
Epoch 10/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 112ms/step - accuracy: 0.9375 - loss: 0.2980 - val_accuracy: 0.9200 - val_loss: 0.4188
Epoch 11/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 87s 3s/step - accuracy: 0.9506 - loss: 0.2925 - val_accuracy: 0.9600 - val_loss: 0.2658
Epoch 12/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 111ms/step - accuracy: 0.9688 - loss: 0.2506 - val_accuracy: 0.9200 - val_loss: 0.3867
Epoch 13/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 87s 3s/step - accuracy: 0.9606 - loss: 0.2647 - val_accuracy: 0.9600 - val_loss: 0.2552
Epoch 14/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 112ms/step - accuracy: 1.0000 - loss: 0.1633 - val_accuracy: 0.9600 - val_loss: 0.2620
Epoch 15/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 89s 3s/step - accuracy: 0.9647 - loss: 0.2232 - val_accuracy: 0.9600 - val_loss: 0.1819
Epoch 16/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 121ms/step - accuracy: 1.0000 - loss: 0.1440 - val_accuracy: 0.9600 - val_loss: 0.2634
Epoch 17/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 87s 3s/step - accuracy: 0.9693 - loss: 0.2086 - val_accuracy: 0.9600 - val_loss: 0.2160
Epoch 18/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 114ms/step - accuracy: 0.9062 - loss: 0.2552 - val_accuracy: 1.0000 - val_loss: 0.2158
Epoch 19/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 87s 3s/step - accuracy: 0.9766 - loss: 0.2131 - val_accuracy: 0.9600 - val_loss: 0.1772
Epoch 20/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 114ms/step - accuracy: 0.9688 - loss: 0.1584 - val_accuracy: 1.0000 - val_loss: 0.1633
Epoch 21/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 88s 3s/step - accuracy: 0.9722 - loss: 0.1967 - val_accuracy: 0.9600 - val_loss: 0.2458
Epoch 22/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 110ms/step - accuracy: 0.9688 - loss: 0.1797 - val_accuracy: 0.8800 - val_loss: 0.3268
Epoch 23/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 87s 3s/step - accuracy: 0.9797 - loss: 0.1902 - val_accuracy: 0.9600 - val_loss: 0.1625
Epoch 24/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 113ms/step - accuracy: 0.9688 - loss: 0.1876 - val_accuracy: 1.0000 - val_loss: 0.1730
Epoch 25/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 87s 3s/step - accuracy: 0.9698 - loss: 0.1968 - val_accuracy: 1.0000 - val_loss: 0.1456
Epoch 26/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 109ms/step - accuracy: 1.0000 - loss: 0.1600 - val_accuracy: 1.0000 - val_loss: 0.1515
Epoch 27/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 89s 3s/step - accuracy: 0.9881 - loss: 0.1693 - val_accuracy: 1.0000 - val_loss: 0.1785
Epoch 28/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 111ms/step - accuracy: 0.9688 - loss: 0.1946 - val_accuracy: 0.9600 - val_loss: 0.1978
Epoch 29/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 87s 3s/step - accuracy: 0.9793 - loss: 0.1947 - val_accuracy: 1.0000 - val_loss: 0.1649
Epoch 30/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 111ms/step - accuracy: 1.0000 - loss: 0.1523 - val_accuracy: 1.0000 - val_loss: 0.1205
Epoch 31/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 87s 3s/step - accuracy: 0.9862 - loss: 0.1702 - val_accuracy: 1.0000 - val_loss: 0.1362
Epoch 32/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 114ms/step - accuracy: 0.9062 - loss: 0.2687 - val_accuracy: 0.9600 - val_loss: 0.2206
Epoch 33/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 87s 3s/step - accuracy: 0.9853 - loss: 0.1688 - val_accuracy: 1.0000 - val_loss: 0.1529
Epoch 34/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 112ms/step - accuracy: 1.0000 - loss: 0.1173 - val_accuracy: 0.9200 - val_loss: 0.2665
Epoch 35/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 88s 3s/step - accuracy: 0.9871 - loss: 0.1554 - val_accuracy: 0.9600 - val_loss: 0.1687
Epoch 36/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 112ms/step - accuracy: 1.0000 - loss: 0.1319 - val_accuracy: 1.0000 - val_loss: 0.1514
Epoch 37/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 88s 3s/step - accuracy: 0.9874 - loss: 0.1497 - val_accuracy: 0.9600 - val_loss: 0.1989
Epoch 38/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 111ms/step - accuracy: 0.9688 - loss: 0.1870 - val_accuracy: 1.0000 - val_loss: 0.1098
Epoch 39/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 88s 3s/step - accuracy: 0.9881 - loss: 0.1524 - val_accuracy: 0.9600 - val_loss: 0.2299
Epoch 40/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 109ms/step - accuracy: 1.0000 - loss: 0.1498 - val_accuracy: 0.9600 - val_loss: 0.2152
Epoch 41/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 88s 3s/step - accuracy: 0.9931 - loss: 0.1487 - val_accuracy: 0.9600 - val_loss: 0.1595
Epoch 42/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 113ms/step - accuracy: 0.9688 - loss: 0.1532 - val_accuracy: 1.0000 - val_loss: 0.1128
Epoch 43/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 90s 3s/step - accuracy: 0.9803 - loss: 0.1550 - val_accuracy: 1.0000 - val_loss: 0.1383
Epoch 44/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 109ms/step - accuracy: 1.0000 - loss: 0.1180 - val_accuracy: 0.9200 - val_loss: 0.2364
Epoch 45/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 88s 3s/step - accuracy: 0.9875 - loss: 0.1447 - val_accuracy: 1.0000 - val_loss: 0.1334
Epoch 46/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 116ms/step - accuracy: 1.0000 - loss: 0.1122 - val_accuracy: 0.9200 - val_loss: 0.2801
Epoch 47/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 88s 3s/step - accuracy: 0.9867 - loss: 0.1586 - val_accuracy: 1.0000 - val_loss: 0.1181
Epoch 48/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 6s 111ms/step - accuracy: 1.0000 - loss: 0.1165 - val_accuracy: 0.9600 - val_loss: 0.1809
Epoch 49/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 90s 3s/step - accuracy: 0.9838 - loss: 0.1422 - val_accuracy: 0.9600 - val_loss: 0.1458
Epoch 50/50
26/26 ━━━━━━━━━━━━━━━━━━━━ 3s 111ms/step - accuracy: 1.0000 - loss: 0.0940 - val_accuracy: 0.9600 - val_loss: 0.1969

import os

# Nombre base del modelo
model_name = "model_VGG16_v"

# Extensión del archivo
file_extension = ".keras"

# Directorio donde se guardarán los modelos
model_directory = "models/"

# Inicializar contador
counter = 1

# Generar el nombre completo del archivo
file_name = model_name + file_extension

ruta=model_directory + file_name
print(ruta)
# Verificar si el modelo ya está guardado
while os.path.exists(model_directory + file_name):
    
    # Si el archivo existe, agregar un número al final del nombre del modelo
    file_name = f"{model_name}{counter}{file_extension}"
    counter += 1

# Guardar el modelo con el nombre único en el directorio correcto
custom_vgg_model.save(model_directory + file_name)

models/model_VGG16_v.keras

def plotTraining(hist, epochs, typeData):

    # Seleccionar la figura y establecer el tamaño
    # Dependiendo del tipo de datos (loss o accuracy), se elige la figura correspondiente
    
    if typeData=="loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'-r',label='Loss Training')
    if typeData=="accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'-r',label='Accuracy Training')
    if typeData=="val_loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['val_loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'--b',label='Loss Validate')
    if typeData=="val_accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['val_accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'--b',label='Training Validate')
        

    plt.rc('xtick',labelsize=24)
    plt.rc('ytick',labelsize=24)
    plt.rc('legend', fontsize=18) 
    plt.legend()
    plt.xlabel('Number of Epochs',fontsize=24)
    plt.grid(True)

plotTraining(model_history,epochs,"loss")
plotTraining(model_history,epochs,"accuracy")
plotTraining(model_history,epochs,"val_loss")
plotTraining(model_history,epochs,"val_accuracy")

from keras.applications.imagenet_utils import preprocess_input, decode_predictions
from keras.models import load_model

names = ['CATHARTES AURA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS']

# Cargar el modelo
modelt = load_model(model_directory + file_name)
#modelt = custom_vgg_model

# Ruta de la imagen de prueba
imaget_path = "ImagenPrueba2.jpg"

# Leer la imagen, cambiar tamaño y preprocesar
imaget=cv2.resize(cv2.imread(imaget_path), (width_shape, height_shape), interpolation = cv2.INTER_AREA)
xt = np.asarray(imaget)
xt=preprocess_input(xt)
xt = np.expand_dims(xt,axis=0)

# Obtener las predicciones del modelo
preds = modelt.predict(xt)

# Obtener la clase predicha y su porcentaje de confianza
predicted_class_index = np.argmax(preds)
predicted_class_name = names[predicted_class_index]
confidence_percentage = preds[0][predicted_class_index] * 100

# Imprimir el resultado
print(f'Clase predicha: {predicted_class_name}')
print(f'Porcentaje de confianza: {confidence_percentage:.2f}%')

# Mostrar la imagen
plt.imshow(cv2.cvtColor(np.asarray(imaget), cv2.COLOR_BGR2RGB))
plt.axis('off')
plt.show()

1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 816ms/step
Clase predicha: CYANOCORAX YNCAS
Porcentaje de confianza: 52.63%

from sklearn.metrics import confusion_matrix, f1_score, roc_curve, precision_score, recall_score, accuracy_score, roc_auc_score
from sklearn import metrics
from mlxtend.plotting import plot_confusion_matrix
from keras.models import load_model
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline


names = ['CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS', 'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS', 'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA', 'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']



test_data_dir = 'dataset/test'  

test_datagen = ImageDataGenerator()

test_generator = test_datagen.flow_from_directory(
    test_data_dir,
    target_size=(width_shape, height_shape), 
    batch_size = batch_size,
    class_mode='categorical', 
    shuffle=False)

custom_Model= load_model(model_directory + file_name)

predictions = custom_Model.predict(test_generator)


y_pred = np.argmax(predictions, axis=1)
y_real = test_generator.classes


matc=confusion_matrix(y_real, y_pred)



print(metrics.classification_report(y_real,y_pred, digits = 4))

Found 27 images belonging to 5 classes.
1/1 ━━━━━━━━━━━━━━━━━━━━ 3s 3s/step
              precision    recall  f1-score   support

           0     1.0000    0.2000    0.3333         5
           1     0.8333    1.0000    0.9091         5
           2     0.7143    1.0000    0.8333         5
           3     0.8333    1.0000    0.9091         5
           4     1.0000    1.0000    1.0000         7

    accuracy                         0.8519        27
   macro avg     0.8762    0.8400    0.7970        27
weighted avg     0.8854    0.8519    0.8120        27

prueba 5

from keras.models import Sequential, Model
from keras.layers import Conv2D, MaxPool2D, Dense, Flatten, Dropout, BatchNormalization, Input
from keras.optimizers import Adam
from keras.callbacks import TensorBoard, ModelCheckpoint
from keras.utils import to_categorical
import os
import numpy as np
from keras.preprocessing import image
from keras.applications.imagenet_utils import preprocess_input, decode_predictions
from keras.applications.vgg16 import VGG16
from tensorflow.keras.preprocessing.image import ImageDataGenerator 
from tensorflow.keras.models import load_model
from sklearn.utils import shuffle
from sklearn.model_selection import train_test_split
import cv2
import matplotlib.pyplot as plt
%matplotlib inline

width_shape = 224
height_shape = 224
num_classes = 18
epochs = 50
batch_size = 32

train_data_dir = 'datasetpreprocesado/train'  
validation_data_dir = 'datasetpreprocesado/valid'

# Definir el generador de imágenes para el conjunto de entrenamiento con aumentos de datos
train_datagen = ImageDataGenerator(  
    rotation_range=20,               # Rango de grados para rotación aleatoria
    zoom_range=0.2,                  # Rango de zoom aleatorio
    width_shift_range=0.1,           # Rango de desplazamiento horizontal aleatorio
    height_shift_range=0.1,          # Rango de desplazamiento vertical aleatorio
    horizontal_flip=True,            # Volteo horizontal aleatorio
    vertical_flip=False,             # No se aplica volteo vertical
    preprocessing_function=preprocess_input)  # Función de preprocesamiento

# Definir el generador de imágenes para el conjunto de validación con los mismos aumentos de datos
valid_datagen = ImageDataGenerator(    
    rotation_range=20,
    zoom_range=0.2,
    width_shift_range=0.1,
    height_shift_range=0.1,
    horizontal_flip=True,
    vertical_flip=False,
    preprocessing_function=preprocess_input)

# Crear un generador de lotes de imágenes para el conjunto de entrenamiento
train_generator = train_datagen.flow_from_directory(  
    train_data_dir,                  # Directorio que contiene las imágenes de entrenamiento
    target_size=(width_shape, height_shape),  # Tamaño al que se redimensionarán las imágenes
    batch_size=batch_size,           # Tamaño del lote
    class_mode='categorical')        # Modo de clasificación para imágenes categóricas

# Crear un generador de lotes de imágenes para el conjunto de validación
validation_generator = valid_datagen.flow_from_directory(  
    validation_data_dir,             # Directorio que contiene las imágenes de validación
    target_size=(width_shape, height_shape),  # Tamaño al que se redimensionarán las imágenes
    batch_size=batch_size,           # Tamaño del lote
    class_mode='categorical')        # Modo de clasificación para imágenes categóricas

Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

# Definir el número de muestras de entrenamiento y validación
nb_train_samples = 2621
nb_validation_samples =  738

# Definir la entrada de la red neuronal con el tamaño de las imágenes
image_input = Input(shape=(width_shape, height_shape, 3))

# Cargar el modelo VGG16 preentrenado con pesos ajustados desde ImageNet
model = VGG16(input_tensor=image_input, include_top=True, weights='imagenet')

# Obtener la salida de la penúltima capa densa del modelo VGG16 (fc2)
last_layer = model.get_layer('fc2').output

# Añadir una nueva capa densa al final del modelo para la clasificación multiclase con regularización L2 (Evita sobreajuste)
out = Dense(num_classes, activation='softmax', kernel_regularizer='l2', name='output')(last_layer)

# Crear un nuevo modelo personalizado que toma la entrada de la imagen y produce la salida clasificada
custom_vgg_model = Model(image_input, out)

# Congelar todas las capas del modelo, excepto la capa densa añadida
for layer in custom_vgg_model.layers[:-1]:
    layer.trainable = False

# Compilar el modelo con una función de pérdida, optimizador y métricas especificadas
custom_vgg_model.compile(loss='categorical_crossentropy', optimizer=Adam(learning_rate=0.0001), metrics=['accuracy'])


# Mostrar un resumen del modelo que incluye la arquitectura y el número de parámetros
custom_vgg_model.summary()

# Entrenar el modelo utilizando generadores de datos para el conjunto de entrenamiento y validación
model_history = custom_vgg_model.fit(
    train_generator,
    epochs=epochs,
    validation_data=validation_generator,
    steps_per_epoch=nb_train_samples//batch_size,  # Número de pasos por época de entrenamiento
    validation_steps=nb_validation_samples//batch_size)  # Número de pasos por época de validación

import os

# Nombre base del modelo
model_name = "model_VGG16_v"

# Extensión del archivo
file_extension = ".keras"

# Directorio donde se guardarán los modelos
model_directory = "models/"

# Inicializar contador
counter = 1

# Generar el nombre completo del archivo
file_name = model_name + file_extension

ruta=model_directory + file_name
print(ruta)
# Verificar si el modelo ya está guardado
while os.path.exists(model_directory + file_name):
    
    # Si el archivo existe, agregar un número al final del nombre del modelo
    file_name = f"{model_name}{counter}{file_extension}"
    counter += 1

# Guardar el modelo con el nombre único en el directorio correcto
custom_vgg_model.save(model_directory + file_name)

models/model_VGG16_v.keras

def plotTraining(hist, epochs, typeData):

    # Seleccionar la figura y establecer el tamaño
    # Dependiendo del tipo de datos (loss o accuracy), se elige la figura correspondiente
    
    if typeData=="loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'-r',label='Loss Training')
    if typeData=="accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'-r',label='Accuracy Training')
    if typeData=="val_loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['val_loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'--b',label='Loss Validate')
    if typeData=="val_accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['val_accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'--b',label='Training Validate')
        

    plt.rc('xtick',labelsize=24)
    plt.rc('ytick',labelsize=24)
    plt.rc('legend', fontsize=18) 
    plt.legend()
    plt.xlabel('Number of Epochs',fontsize=24)
    plt.grid(True)

def plotTraining(hist, epochs, typeData):

    # Seleccionar la figura y establecer el tamaño
    # Dependiendo del tipo de datos (loss o accuracy), se elige la figura correspondiente
    
    if typeData=="loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'-r',label='Loss Training')
    if typeData=="accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'-r',label='Accuracy Training')
    if typeData=="val_loss":
        plt.figure(1,figsize=(10,5))
        yc=hist.history['val_loss']
        xc=range(epochs)
        plt.ylabel('Loss', fontsize=24)
        plt.plot(xc,yc,'--b',label='Loss Validate')
    if typeData=="val_accuracy":
        plt.figure(2,figsize=(10,5))
        yc=hist.history['val_accuracy']
        for i in range(0, len(yc)):
            yc[i]=100*yc[i]
        xc=range(epochs)
        plt.ylabel('Accuracy (%)', fontsize=24)
        plt.plot(xc,yc,'--b',label='Training Validate')
        

    plt.rc('xtick',labelsize=24)
    plt.rc('ytick',labelsize=24)
    plt.rc('legend', fontsize=18) 
    plt.legend()
    plt.xlabel('Number of Epochs',fontsize=24)
    plt.grid(True)

plotTraining(model_history,epochs,"loss")
plotTraining(model_history,epochs,"accuracy")
plotTraining(model_history,epochs,"val_loss")
plotTraining(model_history,epochs,"val_accuracy")

from sklearn.metrics import confusion_matrix, f1_score, roc_curve, precision_score, recall_score, accuracy_score, roc_auc_score
from sklearn import metrics
from mlxtend.plotting import plot_confusion_matrix
from keras.models import load_model
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline


names = ['CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS', 'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS', 'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA', 'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']



test_data_dir = 'datasetpreprocesado/test'  

test_datagen = ImageDataGenerator()

test_generator = test_datagen.flow_from_directory(
    test_data_dir,
    target_size=(width_shape, height_shape), 
    batch_size = batch_size,
    class_mode='categorical', 
    shuffle=False)

custom_Model= load_model('models/model_VGG16_v5.keras')

predictions = custom_Model.predict(test_generator)


y_pred = np.argmax(predictions, axis=1)
y_real = test_generator.classes


matc=confusion_matrix(y_real, y_pred)



print(metrics.classification_report(y_real,y_pred, digits = 4))

Found 738 images belonging to 18 classes.

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\keras\src\trainers\data_adapters\py_dataset_adapter.py:120: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.
  self._warn_if_super_not_called()

24/24 ━━━━━━━━━━━━━━━━━━━━ 76s 3s/step
              precision    recall  f1-score   support

           0     1.0000    0.2927    0.4528        41
           1     0.8095    0.4146    0.5484        41
           2     0.6949    1.0000    0.8200        41
           3     0.7400    0.9024    0.8132        41
           4     0.4933    0.9024    0.6379        41
           5     0.5600    0.6829    0.6154        41
           6     0.6308    1.0000    0.7736        41
           7     0.6271    0.9024    0.7400        41
           8     0.9118    0.7561    0.8267        41
           9     0.2871    0.7073    0.4085        41
          10     0.7750    0.7561    0.7654        41
          11     0.6500    0.6341    0.6420        41
          12     1.0000    0.2683    0.4231        41
          13     0.8000    0.0976    0.1739        41
          14     0.8529    0.7073    0.7733        41
          15     1.0000    0.1951    0.3265        41
          16     0.4091    0.4390    0.4235        41
          17     0.7333    0.5366    0.6197        41

    accuracy                         0.6220       738
   macro avg     0.7208    0.6220    0.5991       738
weighted avg     0.7208    0.6220    0.5991       738

from sklearn.metrics import confusion_matrix, f1_score, roc_curve, precision_score, recall_score, accuracy_score, roc_auc_score
from sklearn import metrics
from mlxtend.plotting import plot_confusion_matrix
from keras.models import load_model
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline


names = [ 'CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS', 'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS', 'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA', 'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']



test_data_dir = 'datasetpreprocesado/test'  

test_datagen = ImageDataGenerator()

test_generator = test_datagen.flow_from_directory(
    test_data_dir,
    target_size=(width_shape, height_shape), 
    batch_size = batch_size,
    class_mode='categorical', 
    shuffle=False)

custom_Model50= load_model('models/model_VGG16_v5.keras')

predictions = custom_Model50.predict(test_generator)


y_pred = np.argmax(predictions, axis=1)
y_real = test_generator.classes


matc=confusion_matrix(y_real, y_pred)

plot_confusion_matrix(conf_mat=matc, figsize=(9,9), class_names = names, show_normed=False)
plt.tight_layout()

print(metrics.classification_report(y_real,y_pred, digits = 4))

Found 361 images belonging to 18 classes.

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\keras\src\trainers\data_adapters\py_dataset_adapter.py:120: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.
  self._warn_if_super_not_called()

5/6 ━━━━━━━━━━━━━━━━━━━━ 6s 7s/step WARNING:tensorflow:5 out of the last 13 calls to <function TensorFlowTrainer.make_predict_function.<locals>.one_step_on_data_distributed at 0x00000153108E7100> triggered tf.function retracing. Tracing is expensive and the excessive number of tracings could be due to (1) creating @tf.function repeatedly in a loop, (2) passing tensors with different shapes, (3) passing Python objects instead of tensors. For (1), please define your @tf.function outside of the loop. For (2), @tf.function has reduce_retracing=True option that can avoid unnecessary retracing. For (3), please refer to https://www.tensorflow.org/guide/function#controlling_retracing and https://www.tensorflow.org/api_docs/python/tf/function for  more details.
6/6 ━━━━━━━━━━━━━━━━━━━━ 40s 6s/step
              precision    recall  f1-score   support

           0     1.0000    0.2000    0.3333        20
           1     0.9167    0.5500    0.6875        20
           2     0.7037    0.9500    0.8085        20
           3     0.5862    0.8500    0.6939        20
           4     0.4146    0.8500    0.5574        20
           5     0.5357    0.7500    0.6250        20
           6     0.5714    1.0000    0.7273        20
           7     0.6296    0.8500    0.7234        20
           8     0.9231    0.6000    0.7273        20
           9     0.3200    0.8000    0.4571        20
          10     0.7222    0.6500    0.6842        20
          11     0.5500    0.5500    0.5500        20
          12     1.0000    0.2000    0.3333        20
          13     0.5000    0.0500    0.0909        20
          14     1.0000    0.8000    0.8889        20
          15     1.0000    0.0500    0.0952        20
          16     0.6316    0.6000    0.6154        20
          17     0.6667    0.4762    0.5556        21

    accuracy                         0.5983       361
   macro avg     0.7040    0.5987    0.5641       361
weighted avg     0.7039    0.5983    0.5641       361

prueba 6

width_shape = 224
height_shape = 224
num_classes = 18
epochs = 50
batch_size = 32

train_data_dir = 'datasetpreprocesado/train'  
validation_data_dir = 'datasetpreprocesado/valid'

# Definir el generador de imágenes para el conjunto de entrenamiento con aumentos de datos
train_datagen = ImageDataGenerator(  
    rotation_range=20,               # Rango de grados para rotación aleatoria
    zoom_range=0.2,                  # Rango de zoom aleatorio
    width_shift_range=0.1,           # Rango de desplazamiento horizontal aleatorio
    height_shift_range=0.1,          # Rango de desplazamiento vertical aleatorio
    horizontal_flip=True,            # Volteo horizontal aleatorio
    vertical_flip=False,             # No se aplica volteo vertical
    preprocessing_function=preprocess_input)  # Función de preprocesamiento

# Definir el generador de imágenes para el conjunto de validación con los mismos aumentos de datos
valid_datagen = ImageDataGenerator(    
    rotation_range=20,
    zoom_range=0.2,
    width_shift_range=0.1,
    height_shift_range=0.1,
    horizontal_flip=True,
    vertical_flip=False,
    preprocessing_function=preprocess_input)

# Crear un generador de lotes de imágenes para el conjunto de entrenamiento
train_generator = train_datagen.flow_from_directory(  
    train_data_dir,                  # Directorio que contiene las imágenes de entrenamiento
    target_size=(width_shape, height_shape),  # Tamaño al que se redimensionarán las imágenes
    batch_size=batch_size,           # Tamaño del lote
    class_mode='categorical')        # Modo de clasificación para imágenes categóricas

# Crear un generador de lotes de imágenes para el conjunto de validación
validation_generator = valid_datagen.flow_from_directory(  
    validation_data_dir,             # Directorio que contiene las imágenes de validación
    target_size=(width_shape, height_shape),  # Tamaño al que se redimensionarán las imágenes
    batch_size=batch_size,           # Tamaño del lote
    class_mode='categorical')        # Modo de clasificación para imágenes categóricas

Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Para aplicar la técnica de ajuste de hiperparámetros, como la tasa de aprendizaje, la regularización y el tamaño del lote, podemos realizar una búsqueda de cuadrícula para encontrar la combinación óptima de estos hiperparámetros que maximice la precisión y la sensibilidad del modelo en el conjunto de validación.

Aquí hay un ejemplo de cómo podríamos ajustar la tasa de aprendizaje y el tamaño del lote utilizando una búsqueda de cuadrícula:

# Importaciones necesarias
import time
import psutil
import numpy as np
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Dense, Flatten
from tensorflow.keras.applications.vgg16 import VGG16, preprocess_input
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping
from tensorflow.keras.preprocessing.image import ImageDataGenerator

# Definir el número de muestras de entrenamiento y validación
nb_train_samples = 2621
nb_validation_samples = 738

# Definir el número de épocas
epochs = 50

# Definir el tamaño de las imágenes
width_shape = 224
height_shape = 224

# Definir el número de clases
num_classes = 18  # Ajustar según el número de clases en tu dataset

# Directorios de datos de entrenamiento y validación
train_data_dir = 'datasetpreprocesado/train'  
validation_data_dir = 'datasetpreprocesado/valid'

# Función para crear y entrenar el modelo
def create_and_train_vgg16_model(learning_rate, l2_regularization, batch_size):
    # Crear generadores de datos con el batch_size proporcionado
    train_datagen = ImageDataGenerator(
        rotation_range=20,
        zoom_range=0.2,
        width_shift_range=0.1,
        height_shift_range=0.1,
        horizontal_flip=True,
        vertical_flip=False,
        preprocessing_function=preprocess_input
    )

    valid_datagen = ImageDataGenerator(
        rotation_range=20,
        zoom_range=0.2,
        width_shift_range=0.1,
        height_shift_range=0.1,
        horizontal_flip=True,
        vertical_flip=False,
        preprocessing_function=preprocess_input
    )

    train_generator = train_datagen.flow_from_directory(
        train_data_dir,
        target_size=(width_shape, height_shape),
        batch_size=batch_size,
        class_mode='categorical'
    )

    validation_generator = valid_datagen.flow_from_directory(
        validation_data_dir,
        target_size=(width_shape, height_shape),
        batch_size=batch_size,
        class_mode='categorical'
    )

    # Definir la entrada de la red neuronal con el tamaño de las imágenes
    image_input = Input(shape=(width_shape, height_shape, 3))

    # Cargar el modelo VGG16 preentrenado con pesos ajustados desde ImageNet
    model = VGG16(input_tensor=image_input, include_top=False, weights='imagenet')

    # Aplanar la salida del VGG16
    x = Flatten()(model.output)

    # Añadir una nueva capa densa al final del modelo para la clasificación multiclase con regularización L2
    out = Dense(num_classes, activation='softmax', kernel_regularizer='l2')(x)

    # Crear un nuevo modelo personalizado que toma la entrada de la imagen y produce la salida clasificada
    custom_vgg_model = Model(inputs=model.input, outputs=out)

    # Congelar todas las capas del modelo base VGG16
    for layer in model.layers:
        layer.trainable = False

    # Compilar el modelo con una función de pérdida, optimizador y métricas especificadas
    custom_vgg_model.compile(loss='categorical_crossentropy', optimizer=Adam(learning_rate=learning_rate), metrics=['accuracy'])

    # Mostrar un resumen del modelo que incluye la arquitectura y el número de parámetros
    custom_vgg_model.summary()

    # Medir el tiempo y el uso de CPU/memoria antes de entrenar
    start_time = time.time()
    start_cpu = psutil.cpu_percent(interval=None)
    start_memory = psutil.virtual_memory().used

    # Crear los callbacks para Early Stopping y guardar el mejor modelo
    checkpoint = ModelCheckpoint('best_model.keras', monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')
    early_stopping = EarlyStopping(monitor='val_accuracy', patience=10, verbose=1, restore_best_weights=True)

    # Entrenar el modelo utilizando generadores de datos para el conjunto de entrenamiento y validación
    model_history = custom_vgg_model.fit(
        train_generator,
        epochs=epochs,
        validation_data=validation_generator,
        steps_per_epoch=nb_train_samples // batch_size,
        validation_steps=nb_validation_samples // batch_size,
        callbacks=[checkpoint, early_stopping]
    )

    # Medir el tiempo y el uso de CPU/memoria después de entrenar
    end_time = time.time()
    end_cpu = psutil.cpu_percent(interval=None)
    end_memory = psutil.virtual_memory().used

    # Calcular métricas de tiempo y uso de recursos
    elapsed_time = end_time - start_time
    cpu_usage = end_cpu - start_cpu
    memory_usage = end_memory - start_memory

    print(f"Tiempo transcurrido para el entrenamiento: {elapsed_time} segundos")
    print(f"Uso de CPU durante el entrenamiento: {cpu_usage}%")
    print(f"Aumento en uso de memoria: {memory_usage / (1024 ** 3)} GB")

    return model_history, elapsed_time, cpu_usage, memory_usage

# Definir rangos de búsqueda para hiperparámetros
learning_rates = [0.0001, 0.0005, 0.001]
l2_regularizations = [0.01, 0.05, 0.1]
batch_sizes = [16, 32, 64]

# Variables para almacenar los mejores hiperparámetros y su rendimiento
best_val_accuracy = 0
best_hyperparams = {}

# Realizar la búsqueda de cuadrícula
for learning_rate in learning_rates:
    for l2_regularization in l2_regularizations:
        for batch_size in batch_sizes:
            # Crear y entrenar el modelo con los hiperparámetros actuales
            model_history, elapsed_time, cpu_usage, memory_usage = create_and_train_vgg16_model(learning_rate, l2_regularization, batch_size)
            
            # Obtener la mejor precisión de validación de esta combinación de hiperparámetros
            val_accuracy = np.max(model_history.history['val_accuracy'])
            
            # Imprimir los resultados
            print(f"Resultados para lr={learning_rate}, l2={l2_regularization}, batch_size={batch_size}:")
            print(f"Tiempo: {elapsed_time} segundos, CPU: {cpu_usage}%, Memoria: {memory_usage / (1024 ** 3)} GB")
            print(f"Precisión de validación: {val_accuracy}")
            
            # Actualizar los mejores hiperparámetros si la precisión de validación mejora
            if val_accuracy > best_val_accuracy:
                best_val_accuracy = val_accuracy
                best_hyperparams = {
                    'learning_rate': learning_rate,
                    'l2_regularization': l2_regularization,
                    'batch_size': batch_size,
                    'val_accuracy': val_accuracy,
                    'elapsed_time': elapsed_time,
                    'cpu_usage': cpu_usage,
                    'memory_usage': memory_usage
                }

# Imprimir los mejores hiperparámetros y su rendimiento
print("Mejores hiperparámetros encontrados:")
print(best_hyperparams)

Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_3"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_1 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_1 (Flatten)             │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_1 (Dense)                 │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\keras\src\trainers\data_adapters\py_dataset_adapter.py:120: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.
  self._warn_if_super_not_called()

163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.3239 - loss: 15.1189
Epoch 1: val_accuracy improved from -inf to 0.76630, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 357s 2s/step - accuracy: 0.3251 - loss: 15.0816 - val_accuracy: 0.7663 - val_loss: 3.0483
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.6875 - loss: 6.4706

C:\Users\Oscar Diaz\anaconda3\Lib\contextlib.py:158: UserWarning: Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches. You may need to use the `.repeat()` function when building your dataset.
  self.gen.throw(typ, value, traceback)

Epoch 2: val_accuracy improved from 0.76630 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.6875 - loss: 6.4706 - val_accuracy: 1.0000 - val_loss: 0.3301
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.8212 - loss: 2.2798
Epoch 3: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.8213 - loss: 2.2797 - val_accuracy: 0.8492 - val_loss: 2.0425
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:15 2s/step - accuracy: 0.8125 - loss: 1.3523
Epoch 4: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.8125 - loss: 1.3523 - val_accuracy: 0.5000 - val_loss: 2.3321
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.8915 - loss: 1.4946
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 344s 2s/step - accuracy: 0.8915 - loss: 1.4945 - val_accuracy: 0.8995 - val_loss: 1.4449
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:27 2s/step - accuracy: 0.9375 - loss: 0.6051
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 0.6051 - val_accuracy: 1.0000 - val_loss: 0.3322
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9061 - loss: 1.1731
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 343s 2s/step - accuracy: 0.9062 - loss: 1.1730 - val_accuracy: 0.9212 - val_loss: 1.2573
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:23 2s/step - accuracy: 0.9375 - loss: 0.6550
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 0.6550 - val_accuracy: 1.0000 - val_loss: 0.3332
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9381 - loss: 0.9227
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 343s 2s/step - accuracy: 0.9381 - loss: 0.9227 - val_accuracy: 0.9239 - val_loss: 1.1713
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 5:03 2s/step - accuracy: 0.9375 - loss: 1.0189
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 1.0189 - val_accuracy: 1.0000 - val_loss: 0.3343
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9698 - loss: 0.6550
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 347s 2s/step - accuracy: 0.9697 - loss: 0.6555 - val_accuracy: 0.9511 - val_loss: 0.8576
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:11 2s/step - accuracy: 0.8750 - loss: 1.2820
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.8750 - loss: 1.2820 - val_accuracy: 1.0000 - val_loss: 0.3340
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2095.2359404563904 segundos
Uso de CPU durante el entrenamiento: 69.80000000000001%
Aumento en uso de memoria: 0.4678535461425781 GB
Resultados para lr=0.0001, l2=0.01, batch_size=16:
Tiempo: 2095.2359404563904 segundos, CPU: 69.80000000000001%, Memoria: 0.4678535461425781 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_5"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_2 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_2 (Flatten)             │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_2 (Dense)                 │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.2720 - loss: 15.9114
Epoch 1: val_accuracy improved from -inf to 0.70788, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 347s 4s/step - accuracy: 0.2740 - loss: 15.8460 - val_accuracy: 0.7079 - val_loss: 4.0938
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:14 3s/step - accuracy: 0.8125 - loss: 1.8199
Epoch 2: val_accuracy improved from 0.70788 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 8ms/step - accuracy: 0.8125 - loss: 1.8199 - val_accuracy: 1.0000 - val_loss: 0.3431
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.7644 - loss: 3.3503
Epoch 3: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 352s 4s/step - accuracy: 0.7647 - loss: 3.3437 - val_accuracy: 0.8274 - val_loss: 2.2608
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:04 3s/step - accuracy: 0.8438 - loss: 4.1966
Epoch 4: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.8438 - loss: 4.1966 - val_accuracy: 1.0000 - val_loss: 0.3452
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.8564 - loss: 1.8617
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 341s 4s/step - accuracy: 0.8565 - loss: 1.8605 - val_accuracy: 0.8777 - val_loss: 1.5264
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.9062 - loss: 1.0022
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.9062 - loss: 1.0022 - val_accuracy: 0.5000 - val_loss: 9.8272
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.8963 - loss: 1.3334
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.8963 - loss: 1.3335 - val_accuracy: 0.8967 - val_loss: 1.5497
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:22 3s/step - accuracy: 0.8438 - loss: 2.6329
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.8438 - loss: 2.6329 - val_accuracy: 1.0000 - val_loss: 0.3432
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9127 - loss: 1.1222
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 341s 4s/step - accuracy: 0.9128 - loss: 1.1220 - val_accuracy: 0.9389 - val_loss: 0.8494
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:01 3s/step - accuracy: 0.9688 - loss: 0.9636
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9688 - loss: 0.9636 - val_accuracy: 1.0000 - val_loss: 0.3432
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9344 - loss: 0.9563
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.9344 - loss: 0.9563 - val_accuracy: 0.9389 - val_loss: 0.9235
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:15 3s/step - accuracy: 0.9688 - loss: 0.6038
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.9688 - loss: 0.6038 - val_accuracy: 0.5000 - val_loss: 14.1933
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2088.1648399829865 segundos
Uso de CPU durante el entrenamiento: 52.4%
Aumento en uso de memoria: 0.7475776672363281 GB
Resultados para lr=0.0001, l2=0.01, batch_size=32:
Tiempo: 2088.1648399829865 segundos, CPU: 52.4%, Memoria: 0.7475776672363281 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_7"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_3 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_3 (Flatten)             │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_3 (Dense)                 │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.1922 - loss: 19.0671
Epoch 1: val_accuracy improved from -inf to 0.63210, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 341s 8s/step - accuracy: 0.1956 - loss: 18.9325 - val_accuracy: 0.6321 - val_loss: 5.9457
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:12 6s/step - accuracy: 0.7031 - loss: 3.8762
Epoch 2: val_accuracy did not improve from 0.63210
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.7031 - loss: 3.8762 - val_accuracy: 0.5294 - val_loss: 7.1882
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.6876 - loss: 4.1880
Epoch 3: val_accuracy improved from 0.63210 to 0.76705, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 342s 8s/step - accuracy: 0.6885 - loss: 4.1744 - val_accuracy: 0.7670 - val_loss: 3.3744
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:05 6s/step - accuracy: 0.8125 - loss: 3.1874
Epoch 4: val_accuracy improved from 0.76705 to 0.79412, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 93ms/step - accuracy: 0.8125 - loss: 3.1874 - val_accuracy: 0.7941 - val_loss: 2.1933
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.7991 - loss: 2.5200
Epoch 5: val_accuracy improved from 0.79412 to 0.83381, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.7996 - loss: 2.5144 - val_accuracy: 0.8338 - val_loss: 2.2552
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.8438 - loss: 2.4263
Epoch 6: val_accuracy improved from 0.83381 to 0.88235, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 102ms/step - accuracy: 0.8438 - loss: 2.4263 - val_accuracy: 0.8824 - val_loss: 2.1459
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.8786 - loss: 1.4324
Epoch 7: val_accuracy did not improve from 0.88235
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.8785 - loss: 1.4341 - val_accuracy: 0.8693 - val_loss: 1.6446
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.7656 - loss: 3.4719
Epoch 8: val_accuracy did not improve from 0.88235
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.7656 - loss: 3.4719 - val_accuracy: 0.8529 - val_loss: 2.0173
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.8951 - loss: 1.4887
Epoch 9: val_accuracy improved from 0.88235 to 0.88636, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 333s 8s/step - accuracy: 0.8952 - loss: 1.4857 - val_accuracy: 0.8864 - val_loss: 1.3530
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:53 6s/step - accuracy: 0.9219 - loss: 0.9825
Epoch 10: val_accuracy improved from 0.88636 to 0.91176, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 101ms/step - accuracy: 0.9219 - loss: 0.9825 - val_accuracy: 0.9118 - val_loss: 1.8430
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9223 - loss: 1.0613
Epoch 11: val_accuracy did not improve from 0.91176
40/40 ━━━━━━━━━━━━━━━━━━━━ 334s 8s/step - accuracy: 0.9222 - loss: 1.0610 - val_accuracy: 0.8991 - val_loss: 1.4575
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:06 6s/step - accuracy: 0.9531 - loss: 0.6821
Epoch 12: val_accuracy improved from 0.91176 to 0.97059, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 92ms/step - accuracy: 0.9531 - loss: 0.6821 - val_accuracy: 0.9706 - val_loss: 1.2731
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9265 - loss: 1.0374
Epoch 13: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 335s 8s/step - accuracy: 0.9266 - loss: 1.0358 - val_accuracy: 0.9091 - val_loss: 1.1663
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:05 6s/step - accuracy: 0.9375 - loss: 0.9251
Epoch 14: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 103ms/step - accuracy: 0.9375 - loss: 0.9251 - val_accuracy: 0.8824 - val_loss: 1.1990
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9472 - loss: 0.7411
Epoch 15: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.9471 - loss: 0.7420 - val_accuracy: 0.9332 - val_loss: 1.0630
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.9219 - loss: 0.8108
Epoch 16: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.9219 - loss: 0.8108 - val_accuracy: 0.9412 - val_loss: 0.7146
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9418 - loss: 0.8265
Epoch 17: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.9419 - loss: 0.8258 - val_accuracy: 0.9261 - val_loss: 0.9895
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.9531 - loss: 0.8481
Epoch 18: val_accuracy improved from 0.97059 to 1.00000, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 91ms/step - accuracy: 0.9531 - loss: 0.8481 - val_accuracy: 1.0000 - val_loss: 0.3511
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9609 - loss: 0.7226
Epoch 19: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 332s 8s/step - accuracy: 0.9609 - loss: 0.7225 - val_accuracy: 0.9361 - val_loss: 1.1457
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:52 6s/step - accuracy: 0.9688 - loss: 0.5045
Epoch 20: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 84ms/step - accuracy: 0.9688 - loss: 0.5045 - val_accuracy: 0.8529 - val_loss: 0.9922
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9587 - loss: 0.7028
Epoch 21: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 332s 8s/step - accuracy: 0.9587 - loss: 0.7022 - val_accuracy: 0.9361 - val_loss: 1.0594
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.9688 - loss: 0.5983
Epoch 22: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 93ms/step - accuracy: 0.9688 - loss: 0.5983 - val_accuracy: 0.9118 - val_loss: 0.9925
Epoch 23/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9646 - loss: 0.5919
Epoch 23: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 329s 8s/step - accuracy: 0.9646 - loss: 0.5918 - val_accuracy: 0.9503 - val_loss: 0.6772
Epoch 24/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:14 7s/step - accuracy: 0.9531 - loss: 0.4978
Epoch 24: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.9531 - loss: 0.4978 - val_accuracy: 0.9706 - val_loss: 0.9520
Epoch 25/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9658 - loss: 0.5349
Epoch 25: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9659 - loss: 0.5353 - val_accuracy: 0.9531 - val_loss: 0.7616
Epoch 26/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 1.0000 - loss: 0.3496
Epoch 26: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 84ms/step - accuracy: 1.0000 - loss: 0.3496 - val_accuracy: 0.9706 - val_loss: 0.5385
Epoch 27/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9776 - loss: 0.4854
Epoch 27: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.9775 - loss: 0.4859 - val_accuracy: 0.9474 - val_loss: 0.8444
Epoch 28/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.9688 - loss: 0.5569
Epoch 28: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.9688 - loss: 0.5569 - val_accuracy: 0.9706 - val_loss: 0.4087
Epoch 28: early stopping
Restoring model weights from the end of the best epoch: 18.
Tiempo transcurrido para el entrenamiento: 4834.694913864136 segundos
Uso de CPU durante el entrenamiento: 36.300000000000004%
Aumento en uso de memoria: 0.8008613586425781 GB
Resultados para lr=0.0001, l2=0.01, batch_size=64:
Tiempo: 4834.694913864136 segundos, CPU: 36.300000000000004%, Memoria: 0.8008613586425781 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_9"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_4 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_4 (Flatten)             │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_4 (Dense)                 │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.3164 - loss: 15.9286
Epoch 1: val_accuracy improved from -inf to 0.77174, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 352s 2s/step - accuracy: 0.3175 - loss: 15.8886 - val_accuracy: 0.7717 - val_loss: 3.0940
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.9375 - loss: 0.9383
Epoch 2: val_accuracy improved from 0.77174 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.9375 - loss: 0.9383 - val_accuracy: 1.0000 - val_loss: 0.3275
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.8018 - loss: 2.7480
Epoch 3: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 347s 2s/step - accuracy: 0.8020 - loss: 2.7454 - val_accuracy: 0.8492 - val_loss: 2.1670
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 0.7500 - loss: 2.1549
Epoch 4: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.7500 - loss: 2.1549 - val_accuracy: 1.0000 - val_loss: 0.3364
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9036 - loss: 1.3884
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 342s 2s/step - accuracy: 0.9036 - loss: 1.3890 - val_accuracy: 0.8845 - val_loss: 1.6293
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 1.0000 - loss: 0.3511
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.3511 - val_accuracy: 1.0000 - val_loss: 0.3293
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9175 - loss: 1.1924
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 343s 2s/step - accuracy: 0.9175 - loss: 1.1918 - val_accuracy: 0.9253 - val_loss: 1.1174
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.6875 - loss: 1.6990
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6875 - loss: 1.6990 - val_accuracy: 1.0000 - val_loss: 0.3300
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9415 - loss: 0.8481
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 341s 2s/step - accuracy: 0.9415 - loss: 0.8484 - val_accuracy: 0.9416 - val_loss: 1.0386
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:18 2s/step - accuracy: 0.9375 - loss: 1.4181
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 1.4181 - val_accuracy: 1.0000 - val_loss: 0.3307
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9387 - loss: 0.9345
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 341s 2s/step - accuracy: 0.9388 - loss: 0.9342 - val_accuracy: 0.9484 - val_loss: 0.7660
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 0.8750 - loss: 0.5370
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.8750 - loss: 0.5370 - val_accuracy: 1.0000 - val_loss: 0.3313
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2077.7854232788086 segundos
Uso de CPU durante el entrenamiento: 1.5%
Aumento en uso de memoria: 0.2444610595703125 GB
Resultados para lr=0.0001, l2=0.05, batch_size=16:
Tiempo: 2077.7854232788086 segundos, CPU: 1.5%, Memoria: 0.2444610595703125 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_11"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_5 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_5 (Flatten)             │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_5 (Dense)                 │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.2749 - loss: 17.9757
Epoch 1: val_accuracy improved from -inf to 0.69565, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 351s 4s/step - accuracy: 0.2771 - loss: 17.8896 - val_accuracy: 0.6957 - val_loss: 3.9034
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:02 3s/step - accuracy: 0.7188 - loss: 5.1091
Epoch 2: val_accuracy improved from 0.69565 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 8ms/step - accuracy: 0.7188 - loss: 5.1091 - val_accuracy: 1.0000 - val_loss: 0.3802
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.7577 - loss: 3.2050
Epoch 3: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 352s 4s/step - accuracy: 0.7580 - loss: 3.1992 - val_accuracy: 0.8098 - val_loss: 2.3410
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:09 3s/step - accuracy: 0.8750 - loss: 2.6156
Epoch 4: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.8750 - loss: 2.6156 - val_accuracy: 1.0000 - val_loss: 0.3446
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.8479 - loss: 1.9547
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.8481 - loss: 1.9520 - val_accuracy: 0.8859 - val_loss: 1.5579
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.9375 - loss: 0.6433
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.9375 - loss: 0.6433 - val_accuracy: 1.0000 - val_loss: 0.3449
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9015 - loss: 1.2463
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 348s 4s/step - accuracy: 0.9015 - loss: 1.2457 - val_accuracy: 0.9090 - val_loss: 1.2508
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:05 3s/step - accuracy: 0.9062 - loss: 1.0065
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.9062 - loss: 1.0065 - val_accuracy: 1.0000 - val_loss: 0.3449
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9299 - loss: 1.0240
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 350s 4s/step - accuracy: 0.9299 - loss: 1.0245 - val_accuracy: 0.9062 - val_loss: 1.3369
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:02 3s/step - accuracy: 0.9375 - loss: 1.0571
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.9375 - loss: 1.0571 - val_accuracy: 1.0000 - val_loss: 0.3448
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9457 - loss: 0.8487
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 349s 4s/step - accuracy: 0.9456 - loss: 0.8494 - val_accuracy: 0.9457 - val_loss: 0.8874
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 5:48 4s/step - accuracy: 0.9062 - loss: 1.3641
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 5s 5ms/step - accuracy: 0.9062 - loss: 1.3641 - val_accuracy: 0.5000 - val_loss: 2.0930
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2118.0271093845367 segundos
Uso de CPU durante el entrenamiento: 25.19999999999999%
Aumento en uso de memoria: -0.6207160949707031 GB
Resultados para lr=0.0001, l2=0.05, batch_size=32:
Tiempo: 2118.0271093845367 segundos, CPU: 25.19999999999999%, Memoria: -0.6207160949707031 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_13"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_6 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_6 (Flatten)             │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_6 (Dense)                 │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.1886 - loss: 19.5179
Epoch 1: val_accuracy improved from -inf to 0.60795, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 345s 8s/step - accuracy: 0.1921 - loss: 19.3752 - val_accuracy: 0.6080 - val_loss: 5.6143
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.5781 - loss: 5.2795
Epoch 2: val_accuracy did not improve from 0.60795
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 99ms/step - accuracy: 0.5781 - loss: 5.2795 - val_accuracy: 0.5588 - val_loss: 6.2684
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.6944 - loss: 4.2246
Epoch 3: val_accuracy improved from 0.60795 to 0.76278, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.6952 - loss: 4.2127 - val_accuracy: 0.7628 - val_loss: 2.8212
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.7500 - loss: 3.9552
Epoch 4: val_accuracy improved from 0.76278 to 0.82353, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 104ms/step - accuracy: 0.7500 - loss: 3.9552 - val_accuracy: 0.8235 - val_loss: 2.4051
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.8005 - loss: 2.4941
Epoch 5: val_accuracy improved from 0.82353 to 0.83949, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 350s 9s/step - accuracy: 0.8010 - loss: 2.4907 - val_accuracy: 0.8395 - val_loss: 2.1663
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:06 6s/step - accuracy: 0.7812 - loss: 2.9902
Epoch 6: val_accuracy did not improve from 0.83949
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 91ms/step - accuracy: 0.7812 - loss: 2.9902 - val_accuracy: 0.7941 - val_loss: 2.3284
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.8697 - loss: 1.6594
Epoch 7: val_accuracy improved from 0.83949 to 0.88778, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 355s 9s/step - accuracy: 0.8699 - loss: 1.6572 - val_accuracy: 0.8878 - val_loss: 1.6590
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:12 6s/step - accuracy: 0.8906 - loss: 1.3438
Epoch 8: val_accuracy did not improve from 0.88778
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 94ms/step - accuracy: 0.8906 - loss: 1.3438 - val_accuracy: 0.8529 - val_loss: 3.2193
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.8963 - loss: 1.3143
Epoch 9: val_accuracy improved from 0.88778 to 0.90341, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.8963 - loss: 1.3148 - val_accuracy: 0.9034 - val_loss: 1.2454
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.8906 - loss: 1.2109
Epoch 10: val_accuracy improved from 0.90341 to 0.97059, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 105ms/step - accuracy: 0.8906 - loss: 1.2109 - val_accuracy: 0.9706 - val_loss: 0.4164
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9270 - loss: 0.9843
Epoch 11: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.9267 - loss: 0.9888 - val_accuracy: 0.9162 - val_loss: 1.1808
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:11 6s/step - accuracy: 0.9844 - loss: 0.4208
Epoch 12: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.9844 - loss: 0.4208 - val_accuracy: 0.9706 - val_loss: 0.4645
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9348 - loss: 0.9616
Epoch 13: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.9347 - loss: 0.9604 - val_accuracy: 0.9304 - val_loss: 0.9586
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:10 6s/step - accuracy: 0.9688 - loss: 0.4415
Epoch 14: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.9688 - loss: 0.4415 - val_accuracy: 0.9118 - val_loss: 1.2959
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9431 - loss: 0.9182
Epoch 15: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.9431 - loss: 0.9169 - val_accuracy: 0.9219 - val_loss: 1.1502
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:54 6s/step - accuracy: 0.9219 - loss: 0.8773
Epoch 16: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 89ms/step - accuracy: 0.9219 - loss: 0.8773 - val_accuracy: 0.8824 - val_loss: 2.1477
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9428 - loss: 0.7785
Epoch 17: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.9428 - loss: 0.7779 - val_accuracy: 0.9375 - val_loss: 0.9925
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.9844 - loss: 0.5313
Epoch 18: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.9844 - loss: 0.5313 - val_accuracy: 0.9118 - val_loss: 2.3550
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9493 - loss: 0.7955
Epoch 19: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 344s 8s/step - accuracy: 0.9494 - loss: 0.7945 - val_accuracy: 0.9403 - val_loss: 0.8397
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.9688 - loss: 0.5107
Epoch 20: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.9688 - loss: 0.5107 - val_accuracy: 0.9118 - val_loss: 0.7869
Epoch 20: early stopping
Restoring model weights from the end of the best epoch: 10.
Tiempo transcurrido para el entrenamiento: 3519.708744764328 segundos
Uso de CPU durante el entrenamiento: 34.2%
Aumento en uso de memoria: 0.21001052856445312 GB
Resultados para lr=0.0001, l2=0.05, batch_size=64:
Tiempo: 3519.708744764328 segundos, CPU: 34.2%, Memoria: 0.21001052856445312 GB
Precisión de validación: 0.970588207244873
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_15"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_7 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_7 (Flatten)             │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_7 (Dense)                 │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.3288 - loss: 14.9486
Epoch 1: val_accuracy improved from -inf to 0.78397, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.3300 - loss: 14.9123 - val_accuracy: 0.7840 - val_loss: 2.6475
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:29 2s/step - accuracy: 0.6875 - loss: 3.8685
Epoch 2: val_accuracy did not improve from 0.78397
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6875 - loss: 3.8685 - val_accuracy: 0.5000 - val_loss: 13.7053
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.8037 - loss: 2.5107
Epoch 3: val_accuracy improved from 0.78397 to 0.86821, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.8038 - loss: 2.5097 - val_accuracy: 0.8682 - val_loss: 1.8337
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:26 2s/step - accuracy: 0.7500 - loss: 3.8194
Epoch 4: val_accuracy improved from 0.86821 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.7500 - loss: 3.8194 - val_accuracy: 1.0000 - val_loss: 0.3340
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.8873 - loss: 1.5738
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.8873 - loss: 1.5739 - val_accuracy: 0.9062 - val_loss: 1.2455
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 5:00 2s/step - accuracy: 0.8125 - loss: 3.1790
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.8125 - loss: 3.1790 - val_accuracy: 1.0000 - val_loss: 0.3335
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9216 - loss: 1.1913
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 348s 2s/step - accuracy: 0.9216 - loss: 1.1910 - val_accuracy: 0.9280 - val_loss: 1.2943
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:22 2s/step - accuracy: 1.0000 - loss: 0.3343
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.3343 - val_accuracy: 1.0000 - val_loss: 0.3343
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9439 - loss: 0.8334
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.9439 - loss: 0.8338 - val_accuracy: 0.9226 - val_loss: 1.2754
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:18 2s/step - accuracy: 1.0000 - loss: 0.3348
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.3348 - val_accuracy: 0.5000 - val_loss: 3.9811
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9411 - loss: 0.9372
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 359s 2s/step - accuracy: 0.9411 - loss: 0.9368 - val_accuracy: 0.9361 - val_loss: 1.0591
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:32 2s/step - accuracy: 1.0000 - loss: 0.3353
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.3353 - val_accuracy: 1.0000 - val_loss: 0.3353
Epoch 13/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9672 - loss: 0.6771
Epoch 13: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.9672 - loss: 0.6773 - val_accuracy: 0.9484 - val_loss: 0.8308
Epoch 14/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.9375 - loss: 1.1467
Epoch 14: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 1.1467 - val_accuracy: 0.5000 - val_loss: 5.1069
Epoch 14: early stopping
Restoring model weights from the end of the best epoch: 4.
Tiempo transcurrido para el entrenamiento: 2501.10480260849 segundos
Uso de CPU durante el entrenamiento: 36.00000000000001%
Aumento en uso de memoria: -0.6569480895996094 GB
Resultados para lr=0.0001, l2=0.1, batch_size=16:
Tiempo: 2501.10480260849 segundos, CPU: 36.00000000000001%, Memoria: -0.6569480895996094 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_17"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_8 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_8 (Flatten)             │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_8 (Dense)                 │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.2459 - loss: 18.2191
Epoch 1: val_accuracy improved from -inf to 0.71467, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.2481 - loss: 18.1370 - val_accuracy: 0.7147 - val_loss: 3.9515
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:20 3s/step - accuracy: 0.6562 - loss: 4.0978
Epoch 2: val_accuracy did not improve from 0.71467
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.6562 - loss: 4.0978 - val_accuracy: 0.5000 - val_loss: 1.6435
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.7742 - loss: 3.2432
Epoch 3: val_accuracy improved from 0.71467 to 0.82609, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.7744 - loss: 3.2406 - val_accuracy: 0.8261 - val_loss: 1.9856
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:13 3s/step - accuracy: 0.8750 - loss: 1.7970
Epoch 4: val_accuracy improved from 0.82609 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 7ms/step - accuracy: 0.8750 - loss: 1.7970 - val_accuracy: 1.0000 - val_loss: 0.3507
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.8541 - loss: 2.0354
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.8543 - loss: 2.0326 - val_accuracy: 0.8682 - val_loss: 1.9640
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:39 3s/step - accuracy: 0.8125 - loss: 1.9255
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.8125 - loss: 1.9255 - val_accuracy: 1.0000 - val_loss: 0.3443
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.8920 - loss: 1.3511
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.8921 - loss: 1.3507 - val_accuracy: 0.9144 - val_loss: 1.1453
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:33 3s/step - accuracy: 0.9062 - loss: 2.5067
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.9062 - loss: 2.5067 - val_accuracy: 1.0000 - val_loss: 0.3443
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9188 - loss: 1.1145
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 349s 4s/step - accuracy: 0.9189 - loss: 1.1132 - val_accuracy: 0.9253 - val_loss: 1.0544
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:27 3s/step - accuracy: 1.0000 - loss: 0.3586
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 1.0000 - loss: 0.3586 - val_accuracy: 1.0000 - val_loss: 0.3443
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9408 - loss: 0.8936
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.9407 - loss: 0.8933 - val_accuracy: 0.9443 - val_loss: 0.9498
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.9688 - loss: 0.5014
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9688 - loss: 0.5014 - val_accuracy: 1.0000 - val_loss: 0.3453
Epoch 13/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9487 - loss: 0.8089
Epoch 13: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.9487 - loss: 0.8086 - val_accuracy: 0.9253 - val_loss: 1.0922
Epoch 14/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:11 3s/step - accuracy: 0.9688 - loss: 0.5556
Epoch 14: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9688 - loss: 0.5556 - val_accuracy: 1.0000 - val_loss: 0.4898
Epoch 14: early stopping
Restoring model weights from the end of the best epoch: 4.
Tiempo transcurrido para el entrenamiento: 2471.026523590088 segundos
Uso de CPU durante el entrenamiento: 58.49999999999999%
Aumento en uso de memoria: 0.4023284912109375 GB
Resultados para lr=0.0001, l2=0.1, batch_size=32:
Tiempo: 2471.026523590088 segundos, CPU: 58.49999999999999%, Memoria: 0.4023284912109375 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_19"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_9 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_9 (Flatten)             │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_9 (Dense)                 │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.1955 - loss: 20.0114
Epoch 1: val_accuracy improved from -inf to 0.61364, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.1991 - loss: 19.8526 - val_accuracy: 0.6136 - val_loss: 5.2234
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.6250 - loss: 5.4743
Epoch 2: val_accuracy did not improve from 0.61364
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.6250 - loss: 5.4743 - val_accuracy: 0.5882 - val_loss: 4.2118
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.7038 - loss: 4.1146
Epoch 3: val_accuracy improved from 0.61364 to 0.74858, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 332s 8s/step - accuracy: 0.7045 - loss: 4.1037 - val_accuracy: 0.7486 - val_loss: 3.3856
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:55 6s/step - accuracy: 0.7500 - loss: 2.9660
Epoch 4: val_accuracy did not improve from 0.74858
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 88ms/step - accuracy: 0.7500 - loss: 2.9660 - val_accuracy: 0.7353 - val_loss: 5.2887
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.8059 - loss: 2.7076
Epoch 5: val_accuracy improved from 0.74858 to 0.82955, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 333s 8s/step - accuracy: 0.8061 - loss: 2.7023 - val_accuracy: 0.8295 - val_loss: 2.3313
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:08 6s/step - accuracy: 0.9062 - loss: 1.0695
Epoch 6: val_accuracy improved from 0.82955 to 0.88235, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 92ms/step - accuracy: 0.9062 - loss: 1.0695 - val_accuracy: 0.8824 - val_loss: 1.6858
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.8726 - loss: 1.6398
Epoch 7: val_accuracy did not improve from 0.88235
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.8727 - loss: 1.6389 - val_accuracy: 0.8679 - val_loss: 1.8297
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.9219 - loss: 1.4198
Epoch 8: val_accuracy did not improve from 0.88235
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 95ms/step - accuracy: 0.9219 - loss: 1.4198 - val_accuracy: 0.8529 - val_loss: 1.1923
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9035 - loss: 1.2833
Epoch 9: val_accuracy improved from 0.88235 to 0.91051, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 348s 9s/step - accuracy: 0.9033 - loss: 1.2849 - val_accuracy: 0.9105 - val_loss: 1.2575
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:13 7s/step - accuracy: 0.9375 - loss: 1.1384
Epoch 10: val_accuracy did not improve from 0.91051
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.9375 - loss: 1.1384 - val_accuracy: 0.8529 - val_loss: 2.2396
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9174 - loss: 1.1176
Epoch 11: val_accuracy did not improve from 0.91051
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.9175 - loss: 1.1172 - val_accuracy: 0.9077 - val_loss: 1.2359
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.9531 - loss: 0.7729
Epoch 12: val_accuracy did not improve from 0.91051
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 96ms/step - accuracy: 0.9531 - loss: 0.7729 - val_accuracy: 0.8235 - val_loss: 1.5957
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9211 - loss: 1.0042
Epoch 13: val_accuracy improved from 0.91051 to 0.92472, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.9212 - loss: 1.0029 - val_accuracy: 0.9247 - val_loss: 0.9894
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:03 6s/step - accuracy: 0.8906 - loss: 1.5692
Epoch 14: val_accuracy did not improve from 0.92472
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.8906 - loss: 1.5692 - val_accuracy: 0.8529 - val_loss: 2.5456
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9359 - loss: 0.8661
Epoch 15: val_accuracy did not improve from 0.92472
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.9360 - loss: 0.8662 - val_accuracy: 0.9190 - val_loss: 1.1437
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:13 7s/step - accuracy: 0.9375 - loss: 0.7110
Epoch 16: val_accuracy improved from 0.92472 to 0.97059, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 97ms/step - accuracy: 0.9375 - loss: 0.7110 - val_accuracy: 0.9706 - val_loss: 0.7691
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9349 - loss: 0.9262
Epoch 17: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.9351 - loss: 0.9235 - val_accuracy: 0.9261 - val_loss: 1.0347
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:24 7s/step - accuracy: 0.9375 - loss: 0.8083
Epoch 18: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 95ms/step - accuracy: 0.9375 - loss: 0.8083 - val_accuracy: 0.9706 - val_loss: 0.5607
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9601 - loss: 0.6419
Epoch 19: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.9599 - loss: 0.6436 - val_accuracy: 0.9389 - val_loss: 0.8187
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.9531 - loss: 0.5130
Epoch 20: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.9531 - loss: 0.5130 - val_accuracy: 0.8824 - val_loss: 1.0313
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9520 - loss: 0.7499
Epoch 21: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.9520 - loss: 0.7493 - val_accuracy: 0.9574 - val_loss: 0.7244
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:10 6s/step - accuracy: 0.9688 - loss: 0.4707
Epoch 22: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.9688 - loss: 0.4707 - val_accuracy: 0.8824 - val_loss: 2.6402
Epoch 23/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9643 - loss: 0.6365
Epoch 23: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 343s 9s/step - accuracy: 0.9642 - loss: 0.6369 - val_accuracy: 0.9361 - val_loss: 1.0553
Epoch 24/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:13 7s/step - accuracy: 0.9531 - loss: 1.1044
Epoch 24: val_accuracy improved from 0.97059 to 1.00000, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 93ms/step - accuracy: 0.9531 - loss: 1.1044 - val_accuracy: 1.0000 - val_loss: 0.3498
Epoch 25/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9611 - loss: 0.7056
Epoch 25: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 332s 8s/step - accuracy: 0.9612 - loss: 0.7053 - val_accuracy: 0.9503 - val_loss: 0.8549
Epoch 26/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.9844 - loss: 0.3862
Epoch 26: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 83ms/step - accuracy: 0.9844 - loss: 0.3862 - val_accuracy: 0.9118 - val_loss: 1.3191
Epoch 27/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9657 - loss: 0.6039
Epoch 27: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 332s 8s/step - accuracy: 0.9658 - loss: 0.6030 - val_accuracy: 0.9545 - val_loss: 0.8867
Epoch 28/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:07 6s/step - accuracy: 0.9688 - loss: 0.4996
Epoch 28: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.9688 - loss: 0.4996 - val_accuracy: 0.9706 - val_loss: 0.4845
Epoch 29/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9643 - loss: 0.5789
Epoch 29: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.9644 - loss: 0.5786 - val_accuracy: 0.9489 - val_loss: 0.8452
Epoch 30/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:55 6s/step - accuracy: 0.9844 - loss: 0.3806
Epoch 30: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 83ms/step - accuracy: 0.9844 - loss: 0.3806 - val_accuracy: 1.0000 - val_loss: 0.3484
Epoch 31/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9774 - loss: 0.5311
Epoch 31: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 333s 8s/step - accuracy: 0.9774 - loss: 0.5315 - val_accuracy: 0.9545 - val_loss: 0.7508
Epoch 32/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:05 6s/step - accuracy: 0.9688 - loss: 0.4833
Epoch 32: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 97ms/step - accuracy: 0.9688 - loss: 0.4833 - val_accuracy: 0.9706 - val_loss: 0.4368
Epoch 33/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9741 - loss: 0.5761
Epoch 33: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.9741 - loss: 0.5749 - val_accuracy: 0.9645 - val_loss: 0.6990
Epoch 34/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.9531 - loss: 0.7288
Epoch 34: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.9531 - loss: 0.7288 - val_accuracy: 0.9412 - val_loss: 0.8989
Epoch 34: early stopping
Restoring model weights from the end of the best epoch: 24.
Tiempo transcurrido para el entrenamiento: 5893.052283287048 segundos
Uso de CPU durante el entrenamiento: 17.799999999999997%
Aumento en uso de memoria: 0.2658805847167969 GB
Resultados para lr=0.0001, l2=0.1, batch_size=64:
Tiempo: 5893.052283287048 segundos, CPU: 17.799999999999997%, Memoria: 0.2658805847167969 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_21"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_10 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_10 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_10 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5546 - loss: 12.8082
Epoch 1: val_accuracy improved from -inf to 0.89538, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 359s 2s/step - accuracy: 0.5557 - loss: 12.7727 - val_accuracy: 0.8954 - val_loss: 2.8618
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:37 2s/step - accuracy: 0.8750 - loss: 1.4841
Epoch 2: val_accuracy did not improve from 0.89538
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.8750 - loss: 1.4841 - val_accuracy: 0.5000 - val_loss: 11.1557
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.8976 - loss: 2.6548
Epoch 3: val_accuracy improved from 0.89538 to 0.92391, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.8977 - loss: 2.6542 - val_accuracy: 0.9239 - val_loss: 2.0025
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:27 2s/step - accuracy: 1.0000 - loss: 0.3994
Epoch 4: val_accuracy improved from 0.92391 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 1.0000 - loss: 0.3994 - val_accuracy: 1.0000 - val_loss: 0.3994
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9292 - loss: 2.0561
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.9292 - loss: 2.0567 - val_accuracy: 0.9416 - val_loss: 2.2820
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:24 2s/step - accuracy: 0.9375 - loss: 0.6514
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 0.6514 - val_accuracy: 1.0000 - val_loss: 0.4584
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9354 - loss: 2.1966
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.9354 - loss: 2.1960 - val_accuracy: 0.9226 - val_loss: 2.8116
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:16 2s/step - accuracy: 0.9375 - loss: 2.3100
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 2.3100 - val_accuracy: 1.0000 - val_loss: 0.5222
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9501 - loss: 1.9485
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 347s 2s/step - accuracy: 0.9501 - loss: 1.9479 - val_accuracy: 0.9361 - val_loss: 2.8034
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:05 2s/step - accuracy: 1.0000 - loss: 0.5833
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.5833 - val_accuracy: 1.0000 - val_loss: 0.5837
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9647 - loss: 1.7192
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 343s 2s/step - accuracy: 0.9647 - loss: 1.7187 - val_accuracy: 0.9484 - val_loss: 2.5617
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:35 2s/step - accuracy: 0.9375 - loss: 2.8202
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 2.8202 - val_accuracy: 1.0000 - val_loss: 0.6315
Epoch 13/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9584 - loss: 2.0239
Epoch 13: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 344s 2s/step - accuracy: 0.9584 - loss: 2.0232 - val_accuracy: 0.9361 - val_loss: 3.0993
Epoch 14/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 1.0000 - loss: 0.6936
Epoch 14: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.6936 - val_accuracy: 1.0000 - val_loss: 0.6940
Epoch 14: early stopping
Restoring model weights from the end of the best epoch: 4.
Tiempo transcurrido para el entrenamiento: 2458.760917901993 segundos
Uso de CPU durante el entrenamiento: 22.60000000000001%
Aumento en uso de memoria: -0.29473114013671875 GB
Resultados para lr=0.0005, l2=0.01, batch_size=16:
Tiempo: 2458.760917901993 segundos, CPU: 22.60000000000001%, Memoria: -0.29473114013671875 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_23"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_11 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_11 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_11 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5144 - loss: 14.6996
Epoch 1: val_accuracy improved from -inf to 0.88179, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 350s 4s/step - accuracy: 0.5167 - loss: 14.6152 - val_accuracy: 0.8818 - val_loss: 2.9613
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:12 3s/step - accuracy: 0.9688 - loss: 1.7583
Epoch 2: val_accuracy improved from 0.88179 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 7ms/step - accuracy: 0.9688 - loss: 1.7583 - val_accuracy: 1.0000 - val_loss: 0.3864
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9136 - loss: 1.9025
Epoch 3: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 346s 4s/step - accuracy: 0.9136 - loss: 1.9027 - val_accuracy: 0.9307 - val_loss: 2.0287
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:08 3s/step - accuracy: 0.8750 - loss: 1.3993
Epoch 4: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.8750 - loss: 1.3993 - val_accuracy: 0.5000 - val_loss: 5.4832
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9349 - loss: 1.3461
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.9348 - loss: 1.3497 - val_accuracy: 0.9280 - val_loss: 1.9616
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:12 3s/step - accuracy: 0.8750 - loss: 1.5736
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.8750 - loss: 1.5736 - val_accuracy: 1.0000 - val_loss: 0.3932
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9467 - loss: 1.3602
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.9467 - loss: 1.3599 - val_accuracy: 0.9606 - val_loss: 1.3207
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:18 3s/step - accuracy: 1.0000 - loss: 0.4216
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 1.0000 - loss: 0.4216 - val_accuracy: 1.0000 - val_loss: 0.4220
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9514 - loss: 1.2499
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.9514 - loss: 1.2513 - val_accuracy: 0.9280 - val_loss: 2.4269
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:05 3s/step - accuracy: 0.9688 - loss: 0.5192
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.9688 - loss: 0.5192 - val_accuracy: 1.0000 - val_loss: 0.4482
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9588 - loss: 1.5175
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.9587 - loss: 1.5186 - val_accuracy: 0.9497 - val_loss: 1.5621
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:15 3s/step - accuracy: 1.0000 - loss: 0.4738
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 1.0000 - loss: 0.4738 - val_accuracy: 1.0000 - val_loss: 0.4744
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2090.6789672374725 segundos
Uso de CPU durante el entrenamiento: 28.900000000000006%
Aumento en uso de memoria: -0.15079116821289062 GB
Resultados para lr=0.0005, l2=0.01, batch_size=32:
Tiempo: 2090.6789672374725 segundos, CPU: 28.900000000000006%, Memoria: -0.15079116821289062 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_25"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_12 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_12 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_12 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4318 - loss: 14.4941
Epoch 1: val_accuracy improved from -inf to 0.86790, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 344s 8s/step - accuracy: 0.4373 - loss: 14.3318 - val_accuracy: 0.8679 - val_loss: 2.4264
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:11 6s/step - accuracy: 0.9062 - loss: 2.3292
Epoch 2: val_accuracy improved from 0.86790 to 0.94118, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 98ms/step - accuracy: 0.9062 - loss: 2.3292 - val_accuracy: 0.9412 - val_loss: 0.5288
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9010 - loss: 2.0859
Epoch 3: val_accuracy did not improve from 0.94118
40/40 ━━━━━━━━━━━━━━━━━━━━ 351s 9s/step - accuracy: 0.9014 - loss: 2.0799 - val_accuracy: 0.9020 - val_loss: 1.8627
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:29 7s/step - accuracy: 0.9531 - loss: 3.0323
Epoch 4: val_accuracy did not improve from 0.94118
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 97ms/step - accuracy: 0.9531 - loss: 3.0323 - val_accuracy: 0.8824 - val_loss: 2.4457
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9297 - loss: 1.5168
Epoch 5: val_accuracy did not improve from 0.94118
40/40 ━━━━━━━━━━━━━━━━━━━━ 333s 8s/step - accuracy: 0.9297 - loss: 1.5194 - val_accuracy: 0.9219 - val_loss: 1.6076
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:08 6s/step - accuracy: 0.9219 - loss: 1.4438
Epoch 6: val_accuracy did not improve from 0.94118
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 84ms/step - accuracy: 0.9219 - loss: 1.4438 - val_accuracy: 0.8529 - val_loss: 2.5131
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9493 - loss: 1.2149
Epoch 7: val_accuracy improved from 0.94118 to 0.94602, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 332s 8s/step - accuracy: 0.9495 - loss: 1.2123 - val_accuracy: 0.9460 - val_loss: 1.4304
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:53 6s/step - accuracy: 0.9688 - loss: 0.9532
Epoch 8: val_accuracy improved from 0.94602 to 0.97059, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 93ms/step - accuracy: 0.9688 - loss: 0.9532 - val_accuracy: 0.9706 - val_loss: 0.6769
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9726 - loss: 0.8028
Epoch 9: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 333s 8s/step - accuracy: 0.9725 - loss: 0.8034 - val_accuracy: 0.9545 - val_loss: 1.0530
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:08 6s/step - accuracy: 0.9844 - loss: 0.7241
Epoch 10: val_accuracy improved from 0.97059 to 1.00000, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 98ms/step - accuracy: 0.9844 - loss: 0.7241 - val_accuracy: 1.0000 - val_loss: 0.4050
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9716 - loss: 0.7662
Epoch 11: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 335s 8s/step - accuracy: 0.9716 - loss: 0.7686 - val_accuracy: 0.9389 - val_loss: 1.3013
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.9375 - loss: 1.3103
Epoch 12: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 86ms/step - accuracy: 0.9375 - loss: 1.3103 - val_accuracy: 0.9706 - val_loss: 2.8014
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9732 - loss: 0.7986
Epoch 13: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9730 - loss: 0.8015 - val_accuracy: 0.9489 - val_loss: 1.3798
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:55 6s/step - accuracy: 0.9688 - loss: 1.5022
Epoch 14: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 95ms/step - accuracy: 0.9688 - loss: 1.5022 - val_accuracy: 0.9706 - val_loss: 1.3243
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9683 - loss: 0.7995
Epoch 15: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.9683 - loss: 0.8012 - val_accuracy: 0.9616 - val_loss: 1.3530
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.9688 - loss: 0.5179
Epoch 16: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 84ms/step - accuracy: 0.9688 - loss: 0.5179 - val_accuracy: 0.9706 - val_loss: 0.4562
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9647 - loss: 0.9873
Epoch 17: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 330s 8s/step - accuracy: 0.9647 - loss: 0.9888 - val_accuracy: 0.9574 - val_loss: 1.4293
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:31 7s/step - accuracy: 0.9219 - loss: 1.1328
Epoch 18: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.9219 - loss: 1.1328 - val_accuracy: 0.9706 - val_loss: 0.7354
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9726 - loss: 0.7761
Epoch 19: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9727 - loss: 0.7782 - val_accuracy: 0.9645 - val_loss: 1.0774
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:11 6s/step - accuracy: 0.9844 - loss: 0.9805
Epoch 20: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.9844 - loss: 0.9805 - val_accuracy: 0.9706 - val_loss: 0.7290
Epoch 20: early stopping
Restoring model weights from the end of the best epoch: 10.
Tiempo transcurrido para el entrenamiento: 3458.426070213318 segundos
Uso de CPU durante el entrenamiento: 30.599999999999994%
Aumento en uso de memoria: 0.24773025512695312 GB
Resultados para lr=0.0005, l2=0.01, batch_size=64:
Tiempo: 3458.426070213318 segundos, CPU: 30.599999999999994%, Memoria: 0.24773025512695312 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_27"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_13 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_13 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_13 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5648 - loss: 13.0838
Epoch 1: val_accuracy improved from -inf to 0.88859, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 344s 2s/step - accuracy: 0.5659 - loss: 13.0474 - val_accuracy: 0.8886 - val_loss: 2.6211
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.7500 - loss: 4.3496
Epoch 2: val_accuracy did not improve from 0.88859
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.7500 - loss: 4.3496 - val_accuracy: 0.5000 - val_loss: 22.2576
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.8735 - loss: 3.6052
Epoch 3: val_accuracy improved from 0.88859 to 0.92391, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 341s 2s/step - accuracy: 0.8735 - loss: 3.6050 - val_accuracy: 0.9239 - val_loss: 2.4035
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 1.0000 - loss: 0.4120
Epoch 4: val_accuracy did not improve from 0.92391
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.4120 - val_accuracy: 0.0000e+00 - val_loss: 3.3323
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9347 - loss: 2.1269
Epoch 5: val_accuracy improved from 0.92391 to 0.93207, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.9347 - loss: 2.1282 - val_accuracy: 0.9321 - val_loss: 2.4189
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:05 2s/step - accuracy: 1.0000 - loss: 0.4679
Epoch 6: val_accuracy improved from 0.93207 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 1.0000 - loss: 0.4679 - val_accuracy: 1.0000 - val_loss: 0.4684
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9403 - loss: 2.0698
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 342s 2s/step - accuracy: 0.9404 - loss: 2.0689 - val_accuracy: 0.9620 - val_loss: 1.5386
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:40 2s/step - accuracy: 1.0000 - loss: 0.5220
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.5220 - val_accuracy: 1.0000 - val_loss: 0.5223
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9553 - loss: 1.6630
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 348s 2s/step - accuracy: 0.9553 - loss: 1.6637 - val_accuracy: 0.9538 - val_loss: 2.3739
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 1.0000 - loss: 0.5682
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.5682 - val_accuracy: 1.0000 - val_loss: 0.5686
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9657 - loss: 1.4156
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.9657 - loss: 1.4152 - val_accuracy: 0.9497 - val_loss: 1.6824
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 1.0000 - loss: 0.6043
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.6043 - val_accuracy: 1.0000 - val_loss: 0.6046
Epoch 13/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9670 - loss: 1.7097
Epoch 13: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 344s 2s/step - accuracy: 0.9670 - loss: 1.7114 - val_accuracy: 0.9389 - val_loss: 2.7864
Epoch 14/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 1.0000 - loss: 0.6567
Epoch 14: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.6567 - val_accuracy: 0.5000 - val_loss: 23.2149
Epoch 15/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9665 - loss: 1.5991
Epoch 15: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 342s 2s/step - accuracy: 0.9665 - loss: 1.5990 - val_accuracy: 0.9606 - val_loss: 1.9011
Epoch 16/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:33 2s/step - accuracy: 1.0000 - loss: 0.7121
Epoch 16: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.7121 - val_accuracy: 1.0000 - val_loss: 0.7124
Epoch 16: early stopping
Restoring model weights from the end of the best epoch: 6.
Tiempo transcurrido para el entrenamiento: 2777.534591436386 segundos
Uso de CPU durante el entrenamiento: 40.7%
Aumento en uso de memoria: 0.08380508422851562 GB
Resultados para lr=0.0005, l2=0.05, batch_size=16:
Tiempo: 2777.534591436386 segundos, CPU: 40.7%, Memoria: 0.08380508422851562 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_29"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_14 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_14 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_14 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.4948 - loss: 12.3616
Epoch 1: val_accuracy improved from -inf to 0.89810, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.4972 - loss: 12.2954 - val_accuracy: 0.8981 - val_loss: 2.1775
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:59 3s/step - accuracy: 0.8125 - loss: 7.0758
Epoch 2: val_accuracy improved from 0.89810 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 8ms/step - accuracy: 0.8125 - loss: 7.0758 - val_accuracy: 1.0000 - val_loss: 0.3430
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9089 - loss: 1.9784
Epoch 3: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 339s 4s/step - accuracy: 0.9089 - loss: 1.9796 - val_accuracy: 0.9212 - val_loss: 1.8589
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:59 3s/step - accuracy: 0.9688 - loss: 0.7362
Epoch 4: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9688 - loss: 0.7362 - val_accuracy: 1.0000 - val_loss: 0.3697
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9381 - loss: 1.4556
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 337s 4s/step - accuracy: 0.9382 - loss: 1.4561 - val_accuracy: 0.9375 - val_loss: 1.9736
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.9375 - loss: 0.9267
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.9375 - loss: 0.9267 - val_accuracy: 1.0000 - val_loss: 0.3942
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9429 - loss: 1.6580
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.9429 - loss: 1.6570 - val_accuracy: 0.9334 - val_loss: 2.3298
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:08 3s/step - accuracy: 1.0000 - loss: 0.4212
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 1.0000 - loss: 0.4212 - val_accuracy: 1.0000 - val_loss: 0.4211
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9507 - loss: 1.5098
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 339s 4s/step - accuracy: 0.9507 - loss: 1.5091 - val_accuracy: 0.9497 - val_loss: 1.5630
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.9688 - loss: 1.9052
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9688 - loss: 1.9052 - val_accuracy: 1.0000 - val_loss: 0.4501
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9595 - loss: 1.2818
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 337s 4s/step - accuracy: 0.9595 - loss: 1.2821 - val_accuracy: 0.9524 - val_loss: 2.0817
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:08 3s/step - accuracy: 0.9688 - loss: 0.5263
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9688 - loss: 0.5263 - val_accuracy: 0.5000 - val_loss: 4.9764
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2059.5964851379395 segundos
Uso de CPU durante el entrenamiento: 22.5%
Aumento en uso de memoria: 0.366241455078125 GB
Resultados para lr=0.0005, l2=0.05, batch_size=32:
Tiempo: 2059.5964851379395 segundos, CPU: 22.5%, Memoria: 0.366241455078125 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_31"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_15 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_15 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_15 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4263 - loss: 14.7463
Epoch 1: val_accuracy improved from -inf to 0.87642, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 334s 8s/step - accuracy: 0.4316 - loss: 14.5857 - val_accuracy: 0.8764 - val_loss: 2.3411
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.8750 - loss: 2.2404
Epoch 2: val_accuracy did not improve from 0.87642
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.8750 - loss: 2.2404 - val_accuracy: 0.8529 - val_loss: 3.7646
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9086 - loss: 1.7937
Epoch 3: val_accuracy improved from 0.87642 to 0.91193, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 335s 8s/step - accuracy: 0.9087 - loss: 1.7930 - val_accuracy: 0.9119 - val_loss: 1.9631
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:55 6s/step - accuracy: 0.8594 - loss: 3.5218
Epoch 4: val_accuracy improved from 0.91193 to 0.97059, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 92ms/step - accuracy: 0.8594 - loss: 3.5218 - val_accuracy: 0.9706 - val_loss: 0.4579
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9438 - loss: 1.3817
Epoch 5: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.9436 - loss: 1.3842 - val_accuracy: 0.9361 - val_loss: 1.4174
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:08 6s/step - accuracy: 0.9531 - loss: 0.6830
Epoch 6: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 84ms/step - accuracy: 0.9531 - loss: 0.6830 - val_accuracy: 0.9412 - val_loss: 1.0407
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9595 - loss: 0.9734
Epoch 7: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.9594 - loss: 0.9744 - val_accuracy: 0.9474 - val_loss: 1.1859
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:07 6s/step - accuracy: 0.9531 - loss: 0.7902
Epoch 8: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.9531 - loss: 0.7902 - val_accuracy: 0.8824 - val_loss: 1.7676
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9644 - loss: 0.8537
Epoch 9: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 330s 8s/step - accuracy: 0.9644 - loss: 0.8559 - val_accuracy: 0.9318 - val_loss: 1.5113
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:07 6s/step - accuracy: 0.9375 - loss: 0.8138
Epoch 10: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 83ms/step - accuracy: 0.9375 - loss: 0.8138 - val_accuracy: 0.9706 - val_loss: 1.2075
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9701 - loss: 0.8206
Epoch 11: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9701 - loss: 0.8213 - val_accuracy: 0.9602 - val_loss: 1.1640
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.9531 - loss: 0.7317
Epoch 12: val_accuracy improved from 0.97059 to 1.00000, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 93ms/step - accuracy: 0.9531 - loss: 0.7317 - val_accuracy: 1.0000 - val_loss: 0.4081
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9721 - loss: 0.8402
Epoch 13: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 330s 8s/step - accuracy: 0.9722 - loss: 0.8397 - val_accuracy: 0.9631 - val_loss: 1.1382
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:53 6s/step - accuracy: 0.9688 - loss: 0.8255
Epoch 14: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 96ms/step - accuracy: 0.9688 - loss: 0.8255 - val_accuracy: 0.9412 - val_loss: 1.2925
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9729 - loss: 0.8347
Epoch 15: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9730 - loss: 0.8342 - val_accuracy: 0.9616 - val_loss: 1.2805
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:06 6s/step - accuracy: 0.9688 - loss: 1.5490
Epoch 16: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.9688 - loss: 1.5490 - val_accuracy: 0.9118 - val_loss: 1.5093
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9682 - loss: 0.9235
Epoch 17: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 330s 8s/step - accuracy: 0.9682 - loss: 0.9233 - val_accuracy: 0.9460 - val_loss: 1.5959
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.9844 - loss: 0.9789
Epoch 18: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.9844 - loss: 0.9789 - val_accuracy: 0.9706 - val_loss: 0.8188
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9721 - loss: 0.9339
Epoch 19: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 333s 8s/step - accuracy: 0.9721 - loss: 0.9333 - val_accuracy: 0.9531 - val_loss: 1.5359
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.9844 - loss: 0.6208
Epoch 20: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 87ms/step - accuracy: 0.9844 - loss: 0.6208 - val_accuracy: 0.9706 - val_loss: 0.5574
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9766 - loss: 0.7381
Epoch 21: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.9766 - loss: 0.7417 - val_accuracy: 0.9659 - val_loss: 1.1822
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.9844 - loss: 0.5217
Epoch 22: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 83ms/step - accuracy: 0.9844 - loss: 0.5217 - val_accuracy: 0.9412 - val_loss: 2.8742
Epoch 22: early stopping
Restoring model weights from the end of the best epoch: 12.
Tiempo transcurrido para el entrenamiento: 3775.056304216385 segundos
Uso de CPU durante el entrenamiento: 36.599999999999994%
Aumento en uso de memoria: 0.05986785888671875 GB
Resultados para lr=0.0005, l2=0.05, batch_size=64:
Tiempo: 3775.056304216385 segundos, CPU: 36.599999999999994%, Memoria: 0.05986785888671875 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_33"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_16 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_16 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_16 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5565 - loss: 12.2703
Epoch 1: val_accuracy improved from -inf to 0.88179, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 344s 2s/step - accuracy: 0.5576 - loss: 12.2386 - val_accuracy: 0.8818 - val_loss: 3.1303
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:07 2s/step - accuracy: 0.9375 - loss: 1.2366
Epoch 2: val_accuracy did not improve from 0.88179
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 1.2366 - val_accuracy: 0.5000 - val_loss: 4.4717
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.8919 - loss: 3.1137
Epoch 3: val_accuracy improved from 0.88179 to 0.89538, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 347s 2s/step - accuracy: 0.8920 - loss: 3.1126 - val_accuracy: 0.8954 - val_loss: 3.1921
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 0.8750 - loss: 2.9106
Epoch 4: val_accuracy improved from 0.89538 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.8750 - loss: 2.9106 - val_accuracy: 1.0000 - val_loss: 0.4032
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9166 - loss: 2.2375
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 343s 2s/step - accuracy: 0.9166 - loss: 2.2376 - val_accuracy: 0.9375 - val_loss: 2.1720
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:12 2s/step - accuracy: 0.8750 - loss: 2.8767
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.8750 - loss: 2.8767 - val_accuracy: 1.0000 - val_loss: 0.4677
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9447 - loss: 1.8778
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 341s 2s/step - accuracy: 0.9447 - loss: 1.8798 - val_accuracy: 0.9212 - val_loss: 3.1020
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:40 2s/step - accuracy: 0.9375 - loss: 7.1626
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 7.1626 - val_accuracy: 1.0000 - val_loss: 0.5284
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9475 - loss: 2.1300
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 341s 2s/step - accuracy: 0.9475 - loss: 2.1295 - val_accuracy: 0.9484 - val_loss: 2.8633
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:02 1s/step - accuracy: 0.9375 - loss: 1.6619
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 1.6619 - val_accuracy: 1.0000 - val_loss: 0.5913
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9633 - loss: 1.4654
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 341s 2s/step - accuracy: 0.9633 - loss: 1.4667 - val_accuracy: 0.9402 - val_loss: 2.9704
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:15 2s/step - accuracy: 1.0000 - loss: 0.6371
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.6371 - val_accuracy: 1.0000 - val_loss: 0.6374
Epoch 13/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9705 - loss: 1.5201
Epoch 13: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.9705 - loss: 1.5198 - val_accuracy: 0.9579 - val_loss: 2.0427
Epoch 14/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 1.0000 - loss: 0.6763
Epoch 14: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.6763 - val_accuracy: 1.0000 - val_loss: 0.6766
Epoch 14: early stopping
Restoring model weights from the end of the best epoch: 4.
Tiempo transcurrido para el entrenamiento: 2419.914089679718 segundos
Uso de CPU durante el entrenamiento: 26.0%
Aumento en uso de memoria: 0.3121757507324219 GB
Resultados para lr=0.0005, l2=0.1, batch_size=16:
Tiempo: 2419.914089679718 segundos, CPU: 26.0%, Memoria: 0.3121757507324219 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_35"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_17 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_17 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_17 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5139 - loss: 13.9244
Epoch 1: val_accuracy improved from -inf to 0.83152, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 338s 4s/step - accuracy: 0.5162 - loss: 13.8457 - val_accuracy: 0.8315 - val_loss: 3.1930
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:58 3s/step - accuracy: 0.9062 - loss: 1.2108
Epoch 2: val_accuracy improved from 0.83152 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 8ms/step - accuracy: 0.9062 - loss: 1.2108 - val_accuracy: 1.0000 - val_loss: 0.3445
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.8895 - loss: 2.4827
Epoch 3: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 339s 4s/step - accuracy: 0.8897 - loss: 2.4774 - val_accuracy: 0.9321 - val_loss: 1.6311
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:00 3s/step - accuracy: 1.0000 - loss: 0.3721
Epoch 4: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 1.0000 - loss: 0.3721 - val_accuracy: 1.0000 - val_loss: 0.3722
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9262 - loss: 1.6709
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 337s 4s/step - accuracy: 0.9262 - loss: 1.6703 - val_accuracy: 0.9185 - val_loss: 1.9659
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:18 3s/step - accuracy: 0.9688 - loss: 0.9096
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.9688 - loss: 0.9096 - val_accuracy: 1.0000 - val_loss: 0.3994
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9504 - loss: 1.3136
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 336s 4s/step - accuracy: 0.9503 - loss: 1.3151 - val_accuracy: 0.9429 - val_loss: 1.7057
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.9375 - loss: 1.2235
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.9375 - loss: 1.2235 - val_accuracy: 1.0000 - val_loss: 0.4265
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9640 - loss: 0.9810
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 339s 4s/step - accuracy: 0.9639 - loss: 0.9830 - val_accuracy: 0.9198 - val_loss: 2.3332
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.9688 - loss: 0.5066
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9688 - loss: 0.5066 - val_accuracy: 1.0000 - val_loss: 0.4471
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9597 - loss: 1.3801
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 350s 4s/step - accuracy: 0.9597 - loss: 1.3791 - val_accuracy: 0.9688 - val_loss: 1.4421
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.9688 - loss: 0.8128
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9688 - loss: 0.8128 - val_accuracy: 1.0000 - val_loss: 0.4746
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2060.525102376938 segundos
Uso de CPU durante el entrenamiento: 24.19999999999999%
Aumento en uso de memoria: -0.09153366088867188 GB
Resultados para lr=0.0005, l2=0.1, batch_size=32:
Tiempo: 2060.525102376938 segundos, CPU: 24.19999999999999%, Memoria: -0.09153366088867188 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_37"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_18 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_18 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_18 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4347 - loss: 15.1208
Epoch 1: val_accuracy improved from -inf to 0.88778, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.4399 - loss: 14.9572 - val_accuracy: 0.8878 - val_loss: 2.3847
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.8594 - loss: 2.2792
Epoch 2: val_accuracy did not improve from 0.88778
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 84ms/step - accuracy: 0.8594 - loss: 2.2792 - val_accuracy: 0.8824 - val_loss: 1.4323
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9011 - loss: 2.0955
Epoch 3: val_accuracy improved from 0.88778 to 0.91477, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 333s 8s/step - accuracy: 0.9014 - loss: 2.0898 - val_accuracy: 0.9148 - val_loss: 1.9255
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:26 7s/step - accuracy: 0.8906 - loss: 2.2663
Epoch 4: val_accuracy improved from 0.91477 to 0.94118, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 92ms/step - accuracy: 0.8906 - loss: 2.2663 - val_accuracy: 0.9412 - val_loss: 3.0590
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9304 - loss: 1.4533
Epoch 5: val_accuracy improved from 0.94118 to 0.95170, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9306 - loss: 1.4521 - val_accuracy: 0.9517 - val_loss: 1.0741
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:05 6s/step - accuracy: 0.9688 - loss: 0.7803
Epoch 6: val_accuracy improved from 0.95170 to 1.00000, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 92ms/step - accuracy: 0.9688 - loss: 0.7803 - val_accuracy: 1.0000 - val_loss: 0.3911
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9454 - loss: 1.1237
Epoch 7: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9455 - loss: 1.1230 - val_accuracy: 0.9460 - val_loss: 1.4682
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.8906 - loss: 2.0263
Epoch 8: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.8906 - loss: 2.0263 - val_accuracy: 0.9118 - val_loss: 1.9532
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9650 - loss: 0.8957
Epoch 9: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 330s 8s/step - accuracy: 0.9649 - loss: 0.8969 - val_accuracy: 0.9602 - val_loss: 1.0955
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.9844 - loss: 0.4297
Epoch 10: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 86ms/step - accuracy: 0.9844 - loss: 0.4297 - val_accuracy: 0.9706 - val_loss: 0.5687
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9703 - loss: 0.9074
Epoch 11: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 330s 8s/step - accuracy: 0.9702 - loss: 0.9082 - val_accuracy: 0.9588 - val_loss: 0.9954
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.9531 - loss: 1.5508
Epoch 12: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 96ms/step - accuracy: 0.9531 - loss: 1.5508 - val_accuracy: 0.9706 - val_loss: 1.9621
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9695 - loss: 0.8586
Epoch 13: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.9695 - loss: 0.8594 - val_accuracy: 0.9588 - val_loss: 1.0309
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:52 6s/step - accuracy: 1.0000 - loss: 0.4111
Epoch 14: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 89ms/step - accuracy: 1.0000 - loss: 0.4111 - val_accuracy: 0.9412 - val_loss: 2.0373
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9755 - loss: 0.8157
Epoch 15: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9755 - loss: 0.8165 - val_accuracy: 0.9588 - val_loss: 1.3062
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:10 6s/step - accuracy: 0.9688 - loss: 1.0216
Epoch 16: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.9688 - loss: 1.0216 - val_accuracy: 0.9412 - val_loss: 2.5329
Epoch 16: early stopping
Restoring model weights from the end of the best epoch: 6.
Tiempo transcurrido para el entrenamiento: 2745.582318544388 segundos
Uso de CPU durante el entrenamiento: 27.099999999999994%
Aumento en uso de memoria: -0.22264480590820312 GB
Resultados para lr=0.0005, l2=0.1, batch_size=64:
Tiempo: 2745.582318544388 segundos, CPU: 27.099999999999994%, Memoria: -0.22264480590820312 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_39"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_19 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_19 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_19 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5745 - loss: 16.4246
Epoch 1: val_accuracy improved from -inf to 0.84647, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 344s 2s/step - accuracy: 0.5755 - loss: 16.3879 - val_accuracy: 0.8465 - val_loss: 6.8197
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:05 2s/step - accuracy: 0.6250 - loss: 20.9407
Epoch 2: val_accuracy improved from 0.84647 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.6250 - loss: 20.9407 - val_accuracy: 1.0000 - val_loss: 0.5724
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9014 - loss: 5.7236
Epoch 3: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 346s 2s/step - accuracy: 0.9014 - loss: 5.7240 - val_accuracy: 0.8967 - val_loss: 6.0108
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:12 2s/step - accuracy: 1.0000 - loss: 0.8505
Epoch 4: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.8505 - val_accuracy: 1.0000 - val_loss: 0.8528
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9311 - loss: 3.8975
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 344s 2s/step - accuracy: 0.9311 - loss: 3.8984 - val_accuracy: 0.9348 - val_loss: 3.6885
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 1.0000 - loss: 1.0928
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 1.0928 - val_accuracy: 1.0000 - val_loss: 1.0945
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9386 - loss: 4.7044
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 341s 2s/step - accuracy: 0.9387 - loss: 4.7015 - val_accuracy: 0.9552 - val_loss: 4.0928
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 1.0000 - loss: 1.3158
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 1.3158 - val_accuracy: 1.0000 - val_loss: 1.3173
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9545 - loss: 3.6700
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 341s 2s/step - accuracy: 0.9545 - loss: 3.6697 - val_accuracy: 0.9552 - val_loss: 4.5430
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:09 2s/step - accuracy: 1.0000 - loss: 1.5116
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 1.5116 - val_accuracy: 1.0000 - val_loss: 1.5125
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9669 - loss: 3.2491
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 348s 2s/step - accuracy: 0.9669 - loss: 3.2527 - val_accuracy: 0.9538 - val_loss: 5.0933
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:20 2s/step - accuracy: 0.9375 - loss: 2.7796
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 2.7796 - val_accuracy: 1.0000 - val_loss: 1.6869
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2076.202100753784 segundos
Uso de CPU durante el entrenamiento: 25.89999999999999%
Aumento en uso de memoria: 0.1414337158203125 GB
Resultados para lr=0.001, l2=0.01, batch_size=16:
Tiempo: 2076.202100753784 segundos, CPU: 25.89999999999999%, Memoria: 0.1414337158203125 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_41"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_20 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_20 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_20 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5123 - loss: 19.5326
Epoch 1: val_accuracy improved from -inf to 0.88723, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 341s 4s/step - accuracy: 0.5147 - loss: 19.4293 - val_accuracy: 0.8872 - val_loss: 4.3339
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:58 3s/step - accuracy: 0.8750 - loss: 4.3965
Epoch 2: val_accuracy improved from 0.88723 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 8ms/step - accuracy: 0.8750 - loss: 4.3965 - val_accuracy: 1.0000 - val_loss: 0.4681
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9111 - loss: 3.7682
Epoch 3: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 336s 4s/step - accuracy: 0.9111 - loss: 3.7639 - val_accuracy: 0.9389 - val_loss: 2.3771
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.9375 - loss: 0.8779
Epoch 4: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.9375 - loss: 0.8779 - val_accuracy: 1.0000 - val_loss: 0.5829
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9512 - loss: 2.0115
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 337s 4s/step - accuracy: 0.9511 - loss: 2.0167 - val_accuracy: 0.9293 - val_loss: 3.4126
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:24 3s/step - accuracy: 0.9688 - loss: 1.5290
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 3ms/step - accuracy: 0.9688 - loss: 1.5290 - val_accuracy: 1.0000 - val_loss: 0.6739
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9360 - loss: 3.1303
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 338s 4s/step - accuracy: 0.9359 - loss: 3.1396 - val_accuracy: 0.9266 - val_loss: 3.4730
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:01 3s/step - accuracy: 0.9688 - loss: 3.0431
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.9688 - loss: 3.0431 - val_accuracy: 1.0000 - val_loss: 0.8117
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9465 - loss: 3.2690
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 336s 4s/step - accuracy: 0.9466 - loss: 3.2643 - val_accuracy: 0.9470 - val_loss: 3.2240
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:57 3s/step - accuracy: 0.9688 - loss: 3.8612
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9688 - loss: 3.8612 - val_accuracy: 1.0000 - val_loss: 0.9113
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9600 - loss: 2.7340
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.9600 - loss: 2.7346 - val_accuracy: 0.9538 - val_loss: 3.1917
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:01 3s/step - accuracy: 0.9375 - loss: 2.2297
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9375 - loss: 2.2297 - val_accuracy: 1.0000 - val_loss: 0.9982
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2053.183504343033 segundos
Uso de CPU durante el entrenamiento: 15.399999999999991%
Aumento en uso de memoria: -0.08095169067382812 GB
Resultados para lr=0.001, l2=0.01, batch_size=32:
Tiempo: 2053.183504343033 segundos, CPU: 15.399999999999991%, Memoria: -0.08095169067382812 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_43"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_21 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_21 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_21 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4979 - loss: 16.8666
Epoch 1: val_accuracy improved from -inf to 0.89773, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.5027 - loss: 16.6990 - val_accuracy: 0.8977 - val_loss: 3.1693
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:03 6s/step - accuracy: 0.9062 - loss: 1.4228
Epoch 2: val_accuracy did not improve from 0.89773
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.9062 - loss: 1.4228 - val_accuracy: 0.8529 - val_loss: 1.7117
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9099 - loss: 2.9109
Epoch 3: val_accuracy improved from 0.89773 to 0.93040, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 334s 8s/step - accuracy: 0.9101 - loss: 2.9086 - val_accuracy: 0.9304 - val_loss: 2.4162
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:54 6s/step - accuracy: 0.9375 - loss: 1.2634
Epoch 4: val_accuracy did not improve from 0.93040
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.9375 - loss: 1.2634 - val_accuracy: 0.9118 - val_loss: 4.5644
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9481 - loss: 1.9929
Epoch 5: val_accuracy improved from 0.93040 to 0.94318, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 332s 8s/step - accuracy: 0.9481 - loss: 1.9937 - val_accuracy: 0.9432 - val_loss: 1.7826
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:53 6s/step - accuracy: 0.9531 - loss: 0.9078
Epoch 6: val_accuracy improved from 0.94318 to 0.97059, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 99ms/step - accuracy: 0.9531 - loss: 0.9078 - val_accuracy: 0.9706 - val_loss: 0.8937
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9522 - loss: 1.7257
Epoch 7: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9521 - loss: 1.7305 - val_accuracy: 0.9347 - val_loss: 2.6764
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:05 6s/step - accuracy: 0.8906 - loss: 2.4168
Epoch 8: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.8906 - loss: 2.4168 - val_accuracy: 0.9412 - val_loss: 1.1705
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9581 - loss: 1.5863
Epoch 9: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9581 - loss: 1.5883 - val_accuracy: 0.9503 - val_loss: 2.3738
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.9688 - loss: 2.5244
Epoch 10: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 83ms/step - accuracy: 0.9688 - loss: 2.5244 - val_accuracy: 0.9706 - val_loss: 1.7108
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9535 - loss: 2.0779
Epoch 11: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 329s 8s/step - accuracy: 0.9536 - loss: 2.0807 - val_accuracy: 0.9460 - val_loss: 2.8556
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:53 6s/step - accuracy: 0.9688 - loss: 1.6969
Epoch 12: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.9688 - loss: 1.6969 - val_accuracy: 0.9412 - val_loss: 2.5582
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9678 - loss: 1.8018
Epoch 13: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.9678 - loss: 1.7993 - val_accuracy: 0.9588 - val_loss: 2.1104
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:08 6s/step - accuracy: 0.9688 - loss: 1.0288
Epoch 14: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 83ms/step - accuracy: 0.9688 - loss: 1.0288 - val_accuracy: 0.9706 - val_loss: 0.9417
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9752 - loss: 1.5477
Epoch 15: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 333s 8s/step - accuracy: 0.9751 - loss: 1.5507 - val_accuracy: 0.9688 - val_loss: 1.7642
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:54 6s/step - accuracy: 0.9688 - loss: 1.2074
Epoch 16: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 86ms/step - accuracy: 0.9688 - loss: 1.2074 - val_accuracy: 0.9706 - val_loss: 0.9391
Epoch 16: early stopping
Restoring model weights from the end of the best epoch: 6.
Tiempo transcurrido para el entrenamiento: 2743.8381311893463 segundos
Uso de CPU durante el entrenamiento: 22.5%
Aumento en uso de memoria: 0.08846664428710938 GB
Resultados para lr=0.001, l2=0.01, batch_size=64:
Tiempo: 2743.8381311893463 segundos, CPU: 22.5%, Memoria: 0.08846664428710938 GB
Precisión de validación: 0.970588207244873
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_45"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_22 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_22 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_22 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5710 - loss: 17.9512
Epoch 1: val_accuracy improved from -inf to 0.90489, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 344s 2s/step - accuracy: 0.5720 - loss: 17.9129 - val_accuracy: 0.9049 - val_loss: 4.4843
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.9375 - loss: 6.4414
Epoch 2: val_accuracy improved from 0.90489 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.9375 - loss: 6.4414 - val_accuracy: 1.0000 - val_loss: 0.5899
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.8966 - loss: 4.5973
Epoch 3: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 340s 2s/step - accuracy: 0.8966 - loss: 4.5965 - val_accuracy: 0.9198 - val_loss: 4.8002
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:33 2s/step - accuracy: 1.0000 - loss: 0.8673
Epoch 4: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 0.8673 - val_accuracy: 0.5000 - val_loss: 4.4055
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9195 - loss: 5.4913
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 348s 2s/step - accuracy: 0.9195 - loss: 5.4941 - val_accuracy: 0.9348 - val_loss: 5.0406
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 0.9375 - loss: 5.2354
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 5.2354 - val_accuracy: 1.0000 - val_loss: 1.1377
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9428 - loss: 4.1708
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 343s 2s/step - accuracy: 0.9428 - loss: 4.1697 - val_accuracy: 0.9416 - val_loss: 5.2362
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 3:19 1s/step - accuracy: 1.0000 - loss: 1.3401
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 1.3401 - val_accuracy: 1.0000 - val_loss: 1.3416
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9442 - loss: 5.0295
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 352s 2s/step - accuracy: 0.9442 - loss: 5.0272 - val_accuracy: 0.9389 - val_loss: 5.0842
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:02 1s/step - accuracy: 0.9375 - loss: 2.7231
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 2.7231 - val_accuracy: 1.0000 - val_loss: 1.5718
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9594 - loss: 4.1690
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 343s 2s/step - accuracy: 0.9595 - loss: 4.1675 - val_accuracy: 0.9389 - val_loss: 4.8885
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:18 2s/step - accuracy: 0.9375 - loss: 3.0526
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 3.0526 - val_accuracy: 1.0000 - val_loss: 1.7289
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2081.577974796295 segundos
Uso de CPU durante el entrenamiento: 23.900000000000006%
Aumento en uso de memoria: -0.043514251708984375 GB
Resultados para lr=0.001, l2=0.05, batch_size=16:
Tiempo: 2081.577974796295 segundos, CPU: 23.900000000000006%, Memoria: -0.043514251708984375 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_47"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_23 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_23 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_23 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5571 - loss: 16.3684
Epoch 1: val_accuracy improved from -inf to 0.87364, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 350s 4s/step - accuracy: 0.5593 - loss: 16.2828 - val_accuracy: 0.8736 - val_loss: 3.9299
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:35 3s/step - accuracy: 0.9375 - loss: 4.3357
Epoch 2: val_accuracy improved from 0.87364 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 8ms/step - accuracy: 0.9375 - loss: 4.3357 - val_accuracy: 1.0000 - val_loss: 0.4603
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9114 - loss: 2.9168
Epoch 3: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.9115 - loss: 2.9171 - val_accuracy: 0.9280 - val_loss: 3.1590
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:09 3s/step - accuracy: 0.9375 - loss: 0.8917
Epoch 4: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9375 - loss: 0.8917 - val_accuracy: 1.0000 - val_loss: 0.5815
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9343 - loss: 2.5511
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 338s 4s/step - accuracy: 0.9343 - loss: 2.5537 - val_accuracy: 0.9253 - val_loss: 3.7120
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:18 3s/step - accuracy: 0.9375 - loss: 2.5516
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9375 - loss: 2.5516 - val_accuracy: 1.0000 - val_loss: 0.7004
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9431 - loss: 2.3326
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 340s 4s/step - accuracy: 0.9431 - loss: 2.3382 - val_accuracy: 0.9321 - val_loss: 4.0687
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:58 3s/step - accuracy: 0.9062 - loss: 3.6672
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9062 - loss: 3.6672 - val_accuracy: 1.0000 - val_loss: 0.8273
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9539 - loss: 2.7641
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 337s 4s/step - accuracy: 0.9539 - loss: 2.7622 - val_accuracy: 0.9484 - val_loss: 2.4021
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:17 3s/step - accuracy: 0.9688 - loss: 1.6910
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.9688 - loss: 1.6910 - val_accuracy: 1.0000 - val_loss: 0.9394
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9681 - loss: 2.0754
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.9680 - loss: 2.0796 - val_accuracy: 0.9538 - val_loss: 3.1890
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:59 3s/step - accuracy: 1.0000 - loss: 1.0324
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 1.0000 - loss: 1.0324 - val_accuracy: 1.0000 - val_loss: 1.0450
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2075.415571451187 segundos
Uso de CPU durante el entrenamiento: 18.80000000000001%
Aumento en uso de memoria: -0.100433349609375 GB
Resultados para lr=0.001, l2=0.05, batch_size=32:
Tiempo: 2075.415571451187 segundos, CPU: 18.80000000000001%, Memoria: -0.100433349609375 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_49"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_24 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_24 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_24 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5064 - loss: 16.5895
Epoch 1: val_accuracy improved from -inf to 0.88778, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 334s 8s/step - accuracy: 0.5113 - loss: 16.4095 - val_accuracy: 0.8878 - val_loss: 3.9406
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.8438 - loss: 9.4449
Epoch 2: val_accuracy improved from 0.88778 to 0.91176, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 97ms/step - accuracy: 0.8438 - loss: 9.4449 - val_accuracy: 0.9118 - val_loss: 4.7573
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9107 - loss: 2.7772
Epoch 3: val_accuracy improved from 0.91176 to 0.96165, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.9110 - loss: 2.7715 - val_accuracy: 0.9616 - val_loss: 1.8672
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.9688 - loss: 1.4099
Epoch 4: val_accuracy did not improve from 0.96165
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.9688 - loss: 1.4099 - val_accuracy: 0.9412 - val_loss: 2.3316
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9456 - loss: 1.8533
Epoch 5: val_accuracy did not improve from 0.96165
40/40 ━━━━━━━━━━━━━━━━━━━━ 334s 8s/step - accuracy: 0.9456 - loss: 1.8573 - val_accuracy: 0.9389 - val_loss: 1.9441
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:11 6s/step - accuracy: 0.9844 - loss: 0.7252
Epoch 6: val_accuracy did not improve from 0.96165
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.9844 - loss: 0.7252 - val_accuracy: 0.9118 - val_loss: 3.1906
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9532 - loss: 1.8113
Epoch 7: val_accuracy did not improve from 0.96165
40/40 ━━━━━━━━━━━━━━━━━━━━ 330s 8s/step - accuracy: 0.9532 - loss: 1.8120 - val_accuracy: 0.9162 - val_loss: 4.0766
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:53 6s/step - accuracy: 0.9062 - loss: 3.9573
Epoch 8: val_accuracy did not improve from 0.96165
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.9062 - loss: 3.9573 - val_accuracy: 0.9118 - val_loss: 4.7475
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9598 - loss: 1.6696
Epoch 9: val_accuracy did not improve from 0.96165
40/40 ━━━━━━━━━━━━━━━━━━━━ 330s 8s/step - accuracy: 0.9598 - loss: 1.6709 - val_accuracy: 0.9361 - val_loss: 2.7426
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:54 6s/step - accuracy: 0.9531 - loss: 2.1353
Epoch 10: val_accuracy did not improve from 0.96165
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 83ms/step - accuracy: 0.9531 - loss: 2.1353 - val_accuracy: 0.8824 - val_loss: 3.9179
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9581 - loss: 1.9871
Epoch 11: val_accuracy did not improve from 0.96165
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9581 - loss: 1.9825 - val_accuracy: 0.9545 - val_loss: 2.4406
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.9688 - loss: 1.6186
Epoch 12: val_accuracy improved from 0.96165 to 0.97059, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 104ms/step - accuracy: 0.9688 - loss: 1.6186 - val_accuracy: 0.9706 - val_loss: 0.7052
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9712 - loss: 1.7417
Epoch 13: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.9712 - loss: 1.7408 - val_accuracy: 0.9361 - val_loss: 3.1923
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.9844 - loss: 0.8070
Epoch 14: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.9844 - loss: 0.8070 - val_accuracy: 0.9412 - val_loss: 6.5906
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9569 - loss: 2.2819
Epoch 15: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9569 - loss: 2.2812 - val_accuracy: 0.9560 - val_loss: 2.8808
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.9688 - loss: 2.0971
Epoch 16: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 91ms/step - accuracy: 0.9688 - loss: 2.0971 - val_accuracy: 0.9118 - val_loss: 4.8347
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9567 - loss: 2.1789
Epoch 17: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9568 - loss: 2.1733 - val_accuracy: 0.9645 - val_loss: 2.3180
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.9688 - loss: 1.7391
Epoch 18: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 84ms/step - accuracy: 0.9688 - loss: 1.7391 - val_accuracy: 0.9412 - val_loss: 2.0214
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9741 - loss: 1.6501
Epoch 19: val_accuracy improved from 0.97059 to 0.97301, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 332s 8s/step - accuracy: 0.9741 - loss: 1.6510 - val_accuracy: 0.9730 - val_loss: 1.8010
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:55 6s/step - accuracy: 1.0000 - loss: 0.8907
Epoch 20: val_accuracy improved from 0.97301 to 1.00000, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 92ms/step - accuracy: 1.0000 - loss: 0.8907 - val_accuracy: 1.0000 - val_loss: 0.8898
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9810 - loss: 1.3003
Epoch 21: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 330s 8s/step - accuracy: 0.9810 - loss: 1.3031 - val_accuracy: 0.9616 - val_loss: 2.9414
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:53 6s/step - accuracy: 1.0000 - loss: 0.9141
Epoch 22: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 86ms/step - accuracy: 1.0000 - loss: 0.9141 - val_accuracy: 1.0000 - val_loss: 0.9146
Epoch 23/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9790 - loss: 1.6879
Epoch 23: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 333s 8s/step - accuracy: 0.9789 - loss: 1.6890 - val_accuracy: 0.9602 - val_loss: 2.6449
Epoch 24/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.9844 - loss: 1.1336
Epoch 24: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 84ms/step - accuracy: 0.9844 - loss: 1.1336 - val_accuracy: 0.9706 - val_loss: 3.6424
Epoch 25/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9828 - loss: 1.4070
Epoch 25: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.9827 - loss: 1.4093 - val_accuracy: 0.9616 - val_loss: 2.6383
Epoch 26/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.9375 - loss: 2.7656
Epoch 26: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 94ms/step - accuracy: 0.9375 - loss: 2.7656 - val_accuracy: 0.9412 - val_loss: 11.0853
Epoch 27/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9825 - loss: 1.5162
Epoch 27: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 333s 8s/step - accuracy: 0.9825 - loss: 1.5164 - val_accuracy: 0.9702 - val_loss: 2.1395
Epoch 28/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:07 6s/step - accuracy: 0.9844 - loss: 1.4639
Epoch 28: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.9844 - loss: 1.4639 - val_accuracy: 0.9706 - val_loss: 1.4039
Epoch 29/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9875 - loss: 1.3647
Epoch 29: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 330s 8s/step - accuracy: 0.9875 - loss: 1.3696 - val_accuracy: 0.9631 - val_loss: 3.2042
Epoch 30/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:55 6s/step - accuracy: 0.9531 - loss: 3.2836
Epoch 30: val_accuracy did not improve from 1.00000
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 83ms/step - accuracy: 0.9531 - loss: 3.2836 - val_accuracy: 0.9412 - val_loss: 3.3419
Epoch 30: early stopping
Restoring model weights from the end of the best epoch: 20.
Tiempo transcurrido para el entrenamiento: 5138.4053819179535 segundos
Uso de CPU durante el entrenamiento: 25.599999999999994%
Aumento en uso de memoria: 0.0103302001953125 GB
Resultados para lr=0.001, l2=0.05, batch_size=64:
Tiempo: 5138.4053819179535 segundos, CPU: 25.599999999999994%, Memoria: 0.0103302001953125 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_51"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_25 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_25 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_25 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5630 - loss: 20.5819
Epoch 1: val_accuracy improved from -inf to 0.89674, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 348s 2s/step - accuracy: 0.5641 - loss: 20.5299 - val_accuracy: 0.8967 - val_loss: 4.6448
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:12 2s/step - accuracy: 0.7500 - loss: 7.4270
Epoch 2: val_accuracy did not improve from 0.89674
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.7500 - loss: 7.4270 - val_accuracy: 0.5000 - val_loss: 1.8399
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9080 - loss: 4.5251
Epoch 3: val_accuracy improved from 0.89674 to 0.91712, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 343s 2s/step - accuracy: 0.9080 - loss: 4.5274 - val_accuracy: 0.9171 - val_loss: 4.5629
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 0.9375 - loss: 6.8902
Epoch 4: val_accuracy improved from 0.91712 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 4ms/step - accuracy: 0.9375 - loss: 6.8902 - val_accuracy: 1.0000 - val_loss: 0.8191
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9271 - loss: 4.4857
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 341s 2s/step - accuracy: 0.9271 - loss: 4.4860 - val_accuracy: 0.9389 - val_loss: 4.0922
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 1.0000 - loss: 1.0671
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 1.0671 - val_accuracy: 1.0000 - val_loss: 1.0684
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9443 - loss: 4.5494
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 341s 2s/step - accuracy: 0.9442 - loss: 4.5488 - val_accuracy: 0.9361 - val_loss: 4.7699
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.8125 - loss: 11.4939
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.8125 - loss: 11.4939 - val_accuracy: 1.0000 - val_loss: 1.2913
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9426 - loss: 4.9243
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.9426 - loss: 4.9253 - val_accuracy: 0.9321 - val_loss: 5.6842
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:02 1s/step - accuracy: 1.0000 - loss: 1.5480
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 1.5480 - val_accuracy: 1.0000 - val_loss: 1.5494
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9523 - loss: 4.2186
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 340s 2s/step - accuracy: 0.9523 - loss: 4.2199 - val_accuracy: 0.9497 - val_loss: 5.8332
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:09 2s/step - accuracy: 1.0000 - loss: 1.7584
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 1.0000 - loss: 1.7584 - val_accuracy: 1.0000 - val_loss: 1.7594
Epoch 13/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.9650 - loss: 4.1459
Epoch 13: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 345s 2s/step - accuracy: 0.9650 - loss: 4.1464 - val_accuracy: 0.9334 - val_loss: 6.2547
Epoch 14/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:15 2s/step - accuracy: 0.9375 - loss: 10.2722
Epoch 14: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.9375 - loss: 10.2722 - val_accuracy: 1.0000 - val_loss: 1.9231
Epoch 14: early stopping
Restoring model weights from the end of the best epoch: 4.
Tiempo transcurrido para el entrenamiento: 2421.348974943161 segundos
Uso de CPU durante el entrenamiento: 25.599999999999994%
Aumento en uso de memoria: 0.30681610107421875 GB
Resultados para lr=0.001, l2=0.1, batch_size=16:
Tiempo: 2421.348974943161 segundos, CPU: 25.599999999999994%, Memoria: 0.30681610107421875 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_53"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_26 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_26 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_26 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5339 - loss: 19.7976
Epoch 1: val_accuracy improved from -inf to 0.89946, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.5362 - loss: 19.6943 - val_accuracy: 0.8995 - val_loss: 3.6957
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:23 3s/step - accuracy: 0.8750 - loss: 3.5391
Epoch 2: val_accuracy improved from 0.89946 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 9ms/step - accuracy: 0.8750 - loss: 3.5391 - val_accuracy: 1.0000 - val_loss: 0.4663
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9058 - loss: 3.7751
Epoch 3: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 339s 4s/step - accuracy: 0.9058 - loss: 3.7760 - val_accuracy: 0.9293 - val_loss: 3.4588
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:01 3s/step - accuracy: 0.9688 - loss: 1.5962
Epoch 4: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.9688 - loss: 1.5962 - val_accuracy: 1.0000 - val_loss: 0.5867
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9281 - loss: 3.3893
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 338s 4s/step - accuracy: 0.9280 - loss: 3.3929 - val_accuracy: 0.9266 - val_loss: 3.8165
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.9062 - loss: 4.0146
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.9062 - loss: 4.0146 - val_accuracy: 1.0000 - val_loss: 0.8362
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9403 - loss: 3.2047
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 336s 4s/step - accuracy: 0.9403 - loss: 3.2016 - val_accuracy: 0.9511 - val_loss: 2.7556
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:26 3s/step - accuracy: 0.9688 - loss: 1.1360
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.9688 - loss: 1.1360 - val_accuracy: 1.0000 - val_loss: 0.8326
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9575 - loss: 2.6649
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.9576 - loss: 2.6622 - val_accuracy: 0.9524 - val_loss: 2.6923
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:58 3s/step - accuracy: 0.8750 - loss: 3.9046
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.8750 - loss: 3.9046 - val_accuracy: 1.0000 - val_loss: 0.9143
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.9702 - loss: 1.9934
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 338s 4s/step - accuracy: 0.9702 - loss: 1.9962 - val_accuracy: 0.9538 - val_loss: 3.2474
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:22 3s/step - accuracy: 0.9375 - loss: 4.3195
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.9375 - loss: 4.3195 - val_accuracy: 1.0000 - val_loss: 0.9902
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2061.667219400406 segundos
Uso de CPU durante el entrenamiento: 25.39999999999999%
Aumento en uso de memoria: -0.2578392028808594 GB
Resultados para lr=0.001, l2=0.1, batch_size=32:
Tiempo: 2061.667219400406 segundos, CPU: 25.39999999999999%, Memoria: -0.2578392028808594 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_55"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_27 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten_27 (Flatten)            │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ dense_27 (Dense)                │ (None, 18)             │       451,602 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 15,166,290 (57.85 MB)

 Trainable params: 451,602 (1.72 MB)

 Non-trainable params: 14,714,688 (56.13 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4760 - loss: 23.8683
Epoch 1: val_accuracy improved from -inf to 0.89915, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.4811 - loss: 23.6032 - val_accuracy: 0.8991 - val_loss: 2.8106
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.9062 - loss: 2.6127
Epoch 2: val_accuracy improved from 0.89915 to 0.91176, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 100ms/step - accuracy: 0.9062 - loss: 2.6127 - val_accuracy: 0.9118 - val_loss: 1.7004
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9219 - loss: 2.2101
Epoch 3: val_accuracy improved from 0.91176 to 0.92188, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 332s 8s/step - accuracy: 0.9219 - loss: 2.2157 - val_accuracy: 0.9219 - val_loss: 2.8222
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:02 6s/step - accuracy: 0.8906 - loss: 1.8345
Epoch 4: val_accuracy did not improve from 0.92188
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.8906 - loss: 1.8345 - val_accuracy: 0.8824 - val_loss: 2.7156
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9514 - loss: 2.1202
Epoch 5: val_accuracy improved from 0.92188 to 0.94460, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 331s 8s/step - accuracy: 0.9514 - loss: 2.1207 - val_accuracy: 0.9446 - val_loss: 1.9858
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.9375 - loss: 3.1698
Epoch 6: val_accuracy improved from 0.94460 to 0.97059, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 92ms/step - accuracy: 0.9375 - loss: 3.1698 - val_accuracy: 0.9706 - val_loss: 1.2765
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9494 - loss: 1.5823
Epoch 7: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 330s 8s/step - accuracy: 0.9493 - loss: 1.5861 - val_accuracy: 0.9432 - val_loss: 2.5169
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.9531 - loss: 2.0113
Epoch 8: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.9531 - loss: 2.0113 - val_accuracy: 0.9412 - val_loss: 0.7040
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.9531 - loss: 1.8427
Epoch 9: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 342s 8s/step - accuracy: 0.9532 - loss: 1.8446 - val_accuracy: 0.9261 - val_loss: 2.9411
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.9062 - loss: 2.0855
Epoch 10: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 84ms/step - accuracy: 0.9062 - loss: 2.0855 - val_accuracy: 0.9412 - val_loss: 1.7005
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9518 - loss: 2.4618
Epoch 11: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 332s 8s/step - accuracy: 0.9520 - loss: 2.4522 - val_accuracy: 0.9560 - val_loss: 2.4249
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:55 6s/step - accuracy: 0.9531 - loss: 1.4318
Epoch 12: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 94ms/step - accuracy: 0.9531 - loss: 1.4318 - val_accuracy: 0.9118 - val_loss: 3.0823
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9598 - loss: 1.7918
Epoch 13: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 332s 8s/step - accuracy: 0.9598 - loss: 1.7926 - val_accuracy: 0.9460 - val_loss: 2.2385
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.9844 - loss: 1.5679
Epoch 14: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.9844 - loss: 1.5679 - val_accuracy: 0.9412 - val_loss: 1.7983
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.9692 - loss: 1.7073
Epoch 15: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 335s 8s/step - accuracy: 0.9692 - loss: 1.7082 - val_accuracy: 0.9389 - val_loss: 2.9604
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 1.0000 - loss: 0.7609
Epoch 16: val_accuracy did not improve from 0.97059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 90ms/step - accuracy: 1.0000 - loss: 0.7609 - val_accuracy: 0.9706 - val_loss: 1.3182
Epoch 16: early stopping
Restoring model weights from the end of the best epoch: 6.
Tiempo transcurrido para el entrenamiento: 2751.2750413417816 segundos
Uso de CPU durante el entrenamiento: 22.799999999999997%
Aumento en uso de memoria: -0.3160667419433594 GB
Resultados para lr=0.001, l2=0.1, batch_size=64:
Tiempo: 2751.2750413417816 segundos, CPU: 22.799999999999997%, Memoria: -0.3160667419433594 GB
Precisión de validación: 0.970588207244873
Mejores hiperparámetros encontrados:
{'learning_rate': 0.0001, 'l2_regularization': 0.01, 'batch_size': 16, 'val_accuracy': 1.0, 'elapsed_time': 2095.2359404563904, 'cpu_usage': 69.80000000000001, 'memory_usage': 502353920}

Explicación de la implementación: Early Stopping:

Se añade el callback EarlyStopping con patience configurado para detener el entrenamiento si no hay mejora en la precisión de validación después de 10 épocas. restore_best_weights=True asegura que el modelo retorne a los pesos de la época con la mejor precisión de validación. ModelCheckpoint:

El callback ModelCheckpoint guarda el mejor modelo basado en la precisión de validación. Seguimiento de métricas:

Se calculan y se imprimen las métricas de tiempo, uso de CPU y memoria para cada combinación de hiperparámetros. Con esta configuración, el entrenamiento se detendrá automáticamente cuando el modelo deje de mejorar, y el mejor modelo encontrado durante la búsqueda de hiperparámetros será guardado.

graficas ultimo modelo

import matplotlib.pyplot as plt

def plotTraining(hist, typeData):
    if typeData == "loss":
        # Obteniendo los valores de la pérdida
        yc = hist.history['loss']
        epochs = len(yc)  # Actualizamos el número de épocas con la longitud real de yc
        xc = range(epochs)
        plt.figure(figsize=(10, 5))
        plt.plot(xc, yc, '-r', label='Pérdida del entrenamiento')
        plt.title('Pérdida del entrenamiento por época')
        plt.xlabel('Épocas', fontsize=14)
        plt.ylabel('Pérdida', fontsize=14)
        plt.legend()
        plt.grid()
        plt.show()

    elif typeData == "accuracy":
        # Obteniendo los valores de la precisión
        yc = hist.history['accuracy']
        epochs = len(yc)  # Actualizamos el número de épocas con la longitud real de yc
        xc = range(epochs)
        plt.figure(figsize=(10, 5))
        plt.plot(xc, yc, '-b', label='Precisión del entrenamiento')
        plt.title('Precisión del entrenamiento por época')
        plt.xlabel('Épocas', fontsize=14)
        plt.ylabel('Precisión', fontsize=14)
        plt.legend()
        plt.grid()
        plt.show()

    elif typeData == "val_loss":
        # Obteniendo los valores de la pérdida de validación
        yc = hist.history['val_loss']
        epochs = len(yc)  # Actualizamos el número de épocas con la longitud real de yc
        xc = range(epochs)
        plt.figure(figsize=(10, 5))
        plt.plot(xc, yc, '-r', label='Pérdida de validación')
        plt.title('Pérdida de validación por época')
        plt.xlabel('Épocas', fontsize=14)
        plt.ylabel('Pérdida', fontsize=14)
        plt.legend()
        plt.grid()
        plt.show()

    elif typeData == "val_accuracy":
        # Obteniendo los valores de la precisión de validación
        yc = hist.history['val_accuracy']
        epochs = len(yc)  # Actualizamos el número de épocas con la longitud real de yc
        xc = range(epochs)
        plt.figure(figsize=(10, 5))
        plt.plot(xc, yc, '-b', label='Precisión de validación')
        plt.title('Precisión de validación por época')
        plt.xlabel('Épocas', fontsize=14)
        plt.ylabel('Precisión', fontsize=14)
        plt.legend()
        plt.grid()
        plt.show()

    else:
        raise ValueError(f"Tipo de dato desconocido: {typeData}. Use 'loss', 'accuracy', 'val_loss' o 'val_accuracy'.")

# Uso de la función para graficar las diferentes métricas
plotTraining(model_history, "loss")
plotTraining(model_history, "accuracy")
plotTraining(model_history, "val_loss")
plotTraining(model_history, "val_accuracy")

import matplotlib.pyplot as plt

def plotTraining(hist):
    epochs = range(len(hist.history['loss']))
    
    plt.figure(figsize=(12, 10))

    # Gráfico de la pérdida
    plt.subplot(2, 1, 1)
    plt.plot(epochs, hist.history['loss'], '-r', label='Pérdida del entrenamiento')
    plt.plot(epochs, hist.history['val_loss'], '-b', label='Pérdida de validación')
    plt.title('Pérdida del entrenamiento y validación')
    plt.xlabel('Épocas', fontsize=14)
    plt.ylabel('Pérdida', fontsize=14)
    plt.legend()
    plt.grid()

    # Gráfico de la precisión
    plt.subplot(2, 1, 2)
    plt.plot(epochs, hist.history['accuracy'], '-r', label='Precisión del entrenamiento')
    plt.plot(epochs, hist.history['val_accuracy'], '-b', label='Precisión de validación')
    plt.title('Precisión del entrenamiento y validación')
    plt.xlabel('Épocas', fontsize=14)
    plt.ylabel('Precisión', fontsize=14)
    plt.legend()
    plt.grid()

    plt.tight_layout()
    plt.show()

# Uso de la función para graficar
plotTraining(model_history)

from sklearn.metrics import confusion_matrix, f1_score, roc_curve, precision_score, recall_score, accuracy_score, roc_auc_score
from sklearn import metrics
from mlxtend.plotting import plot_confusion_matrix
from keras.models import load_model
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline


names = [ 'CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS', 'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS', 'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA', 'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']



test_data_dir = 'datasetpreprocesado/test'  

test_datagen = ImageDataGenerator()

test_generator = test_datagen.flow_from_directory(
    test_data_dir,
    target_size=(width_shape, height_shape), 
    batch_size = batch_size,
    class_mode='categorical', 
    shuffle=False)

custom_Model50= load_model('best_model.keras')

predictions = custom_Model50.predict(test_generator)


y_pred = np.argmax(predictions, axis=1)
y_real = test_generator.classes


matc=confusion_matrix(y_real, y_pred)

plot_confusion_matrix(conf_mat=matc, figsize=(9,9), class_names = names, show_normed=False)
plt.tight_layout()

print(metrics.classification_report(y_real,y_pred, digits = 4))

Found 361 images belonging to 18 classes.

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\keras\src\trainers\data_adapters\py_dataset_adapter.py:120: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.
  self._warn_if_super_not_called()

6/6 ━━━━━━━━━━━━━━━━━━━━ 43s 6s/step
              precision    recall  f1-score   support

           0     1.0000    0.5500    0.7097        20
           1     0.7778    0.3500    0.4828        20
           2     0.8696    1.0000    0.9302        20
           3     0.7083    0.8500    0.7727        20
           4     0.5714    0.8000    0.6667        20
           5     0.3830    0.9000    0.5373        20
           6     0.6897    1.0000    0.8163        20
           7     0.9375    0.7500    0.8333        20
           8     0.9500    0.9500    0.9500        20
           9     0.6250    1.0000    0.7692        20
          10     0.7500    0.9000    0.8182        20
          11     0.8571    0.6000    0.7059        20
          12     0.8000    0.4000    0.5333        20
          13     0.8000    0.2000    0.3200        20
          14     0.9000    0.9000    0.9000        20
          15     1.0000    0.5500    0.7097        20
          16     0.8462    0.5500    0.6667        20
          17     0.4800    0.5714    0.5217        21

    accuracy                         0.7119       361
   macro avg     0.7748    0.7123    0.7024       361
weighted avg     0.7739    0.7119    0.7019       361

Prueba 7

###

import time
import psutil
import numpy as np
from tensorflow.keras.models import Model
from tensorflow.keras.layers import Input, Dense, Flatten
from tensorflow.keras.applications.vgg16 import VGG16, preprocess_input
from tensorflow.keras.optimizers import Adam
from tensorflow.keras.callbacks import ModelCheckpoint, EarlyStopping
from tensorflow.keras.preprocessing.image import ImageDataGenerator

# Definir el número de muestras de entrenamiento y validación
nb_train_samples = 2621
nb_validation_samples = 738

# Definir el número de épocas
epochs = 50

# Definir el tamaño de las imágenes
width_shape = 224
height_shape = 224

# Definir el número de clases
num_classes = 18  # Ajustar según el número de clases en tu dataset

# Directorios de datos de entrenamiento y validación
train_data_dir = 'datasetpreprocesado/train'  
validation_data_dir = 'datasetpreprocesado/valid'

# Función para crear y entrenar el modelo
def create_and_train_vgg16_model(learning_rate, l2_regularization, batch_size):
    # Crear generadores de datos con el batch_size proporcionado
    train_datagen = ImageDataGenerator(
        rotation_range=20,
        zoom_range=0.2,
        width_shift_range=0.1,
        height_shift_range=0.1,
        horizontal_flip=True,
        vertical_flip=False,
        preprocessing_function=preprocess_input
    )

    valid_datagen = ImageDataGenerator(
        rotation_range=20,
        zoom_range=0.2,
        width_shift_range=0.1,
        height_shift_range=0.1,
        horizontal_flip=True,
        vertical_flip=False,
        preprocessing_function=preprocess_input
    )

    train_generator = train_datagen.flow_from_directory(
        train_data_dir,
        target_size=(width_shape, height_shape),
        batch_size=batch_size,
        class_mode='categorical'
    )

    validation_generator = valid_datagen.flow_from_directory(
        validation_data_dir,
        target_size=(width_shape, height_shape),
        batch_size=batch_size,
        class_mode='categorical'
    )

    # Definir la entrada de la red neuronal con el tamaño de las imágenes
    image_input = Input(shape=(width_shape, height_shape, 3))

    # Cargar el modelo VGG16 preentrenado con pesos ajustados desde ImageNet
    base_model = VGG16(input_tensor=image_input, include_top=True, weights='imagenet')

    # Aplanar la salida del VGG16
    x = Flatten(name='custom_flatten')(base_model.output)

    # Añadir una nueva capa densa al final del modelo para la clasificación multiclase con regularización L2
    out = Dense(num_classes, activation='softmax', kernel_regularizer='l2', name='custom_dense')(x)

    # Crear un nuevo modelo personalizado que toma la entrada de la imagen y produce la salida clasificada
    custom_vgg_model = Model(inputs=base_model.input, outputs=out)

    # Congelar todas las capas del modelo base VGG16
    for layer in base_model.layers:
        layer.trainable = False

    # Compilar el modelo con una función de pérdida, optimizador y métricas especificadas
    custom_vgg_model.compile(loss='categorical_crossentropy', optimizer=Adam(learning_rate=learning_rate), metrics=['accuracy'])

    # Mostrar un resumen del modelo que incluye la arquitectura y el número de parámetros
    custom_vgg_model.summary()

    # Medir el tiempo y el uso de CPU/memoria antes de entrenar
    start_time = time.time()
    start_cpu = psutil.cpu_percent(interval=None)
    start_memory = psutil.virtual_memory().used

    # Crear los callbacks para Early Stopping y guardar el mejor modelo
    checkpoint = ModelCheckpoint('best_model.keras', monitor='val_accuracy', verbose=1, save_best_only=True, mode='max')
    early_stopping = EarlyStopping(monitor='val_accuracy', patience=10, verbose=1, restore_best_weights=True)

    # Entrenar el modelo utilizando generadores de datos para el conjunto de entrenamiento y validación
    model_history = custom_vgg_model.fit(
        train_generator,
        epochs=epochs,
        validation_data=validation_generator,
        steps_per_epoch=nb_train_samples // batch_size,
        validation_steps=nb_validation_samples // batch_size,
        callbacks=[checkpoint, early_stopping]
    )

    # Medir el tiempo y el uso de CPU/memoria después de entrenar
    end_time = time.time()
    end_cpu = psutil.cpu_percent(interval=None)
    end_memory = psutil.virtual_memory().used

    # Calcular métricas de tiempo y uso de recursos
    elapsed_time = end_time - start_time
    cpu_usage = end_cpu - start_cpu
    memory_usage = end_memory - start_memory

    print(f"Tiempo transcurrido para el entrenamiento: {elapsed_time} segundos")
    print(f"Uso de CPU durante el entrenamiento: {cpu_usage}%")
    print(f"Aumento en uso de memoria: {memory_usage / (1024 ** 3)} GB")

    return model_history, elapsed_time, cpu_usage, memory_usage

# Definir rangos de búsqueda para hiperparámetros
learning_rates = [0.0001, 0.0005, 0.001]
l2_regularizations = [0.01, 0.05, 0.1]
batch_sizes = [16, 32, 64]

# Variables para almacenar los mejores hiperparámetros y su rendimiento
best_val_accuracy = 0
best_hyperparams = {}

# Realizar la búsqueda de cuadrícula
for learning_rate in learning_rates:
    for l2_regularization in l2_regularizations:
        for batch_size in batch_sizes:
            # Crear y entrenar el modelo con los hiperparámetros actuales
            model_history, elapsed_time, cpu_usage, memory_usage = create_and_train_vgg16_model(learning_rate, l2_regularization, batch_size)
            
            # Obtener la mejor precisión de validación de esta combinación de hiperparámetros
            val_accuracy = np.max(model_history.history['val_accuracy'])
            
            # Imprimir los resultados
            print(f"Resultados para lr={learning_rate}, l2={l2_regularization}, batch_size={batch_size}:")
            print(f"Tiempo: {elapsed_time} segundos, CPU: {cpu_usage}%, Memoria: {memory_usage / (1024 ** 3)} GB")
            print(f"Precisión de validación: {val_accuracy}")
            
            # Actualizar los mejores hiperparámetros si la precisión de validación mejora
            if val_accuracy > best_val_accuracy:
                best_val_accuracy = val_accuracy
                best_hyperparams = {
                    'learning_rate': learning_rate,
                    'l2_regularization': l2_regularization,
                    'batch_size': batch_size,
                    'val_accuracy': val_accuracy,
                    'elapsed_time': elapsed_time,
                    'cpu_usage': cpu_usage,
                    'memory_usage': memory_usage
                }

# Imprimir los mejores hiperparámetros y su rendimiento
print("Mejores hiperparámetros encontrados:")
print(best_hyperparams)

Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_4"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_2 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.0766 - loss: 3.1955
Epoch 1: val_accuracy improved from -inf to 0.13451, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 365s 2s/step - accuracy: 0.0768 - loss: 3.1952 - val_accuracy: 0.1345 - val_loss: 3.0679
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:24 2s/step - accuracy: 0.0625 - loss: 3.0668

C:\Users\Oscar Diaz\anaconda3\Lib\contextlib.py:158: UserWarning: Your input ran out of data; interrupting training. Make sure that your dataset or generator can generate at least `steps_per_epoch * epochs` batches. You may need to use the `.repeat()` function when building your dataset.
  self.gen.throw(typ, value, traceback)

Epoch 2: val_accuracy did not improve from 0.13451
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.0625 - loss: 3.0668 - val_accuracy: 0.0000e+00 - val_loss: 3.0574
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.1637 - loss: 3.0388
Epoch 3: val_accuracy improved from 0.13451 to 0.21467, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 361s 2s/step - accuracy: 0.1638 - loss: 3.0387 - val_accuracy: 0.2147 - val_loss: 2.9675
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:15 2s/step - accuracy: 0.3125 - loss: 2.9683
Epoch 4: val_accuracy did not improve from 0.21467
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3125 - loss: 2.9683 - val_accuracy: 0.0000e+00 - val_loss: 2.9687
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.2599 - loss: 2.9509
Epoch 5: val_accuracy improved from 0.21467 to 0.34239, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.2601 - loss: 2.9508 - val_accuracy: 0.3424 - val_loss: 2.9119
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:25 2s/step - accuracy: 0.5000 - loss: 2.9020
Epoch 6: val_accuracy did not improve from 0.34239
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5000 - loss: 2.9020 - val_accuracy: 0.0000e+00 - val_loss: 2.9371
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.3587 - loss: 2.9014
Epoch 7: val_accuracy improved from 0.34239 to 0.40625, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 353s 2s/step - accuracy: 0.3588 - loss: 2.9014 - val_accuracy: 0.4062 - val_loss: 2.8794
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:15 2s/step - accuracy: 0.6250 - loss: 2.8735
Epoch 8: val_accuracy improved from 0.40625 to 0.50000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 4s 16ms/step - accuracy: 0.6250 - loss: 2.8735 - val_accuracy: 0.5000 - val_loss: 2.8610
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4378 - loss: 2.8739
Epoch 9: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.4378 - loss: 2.8739 - val_accuracy: 0.4660 - val_loss: 2.8605
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.3750 - loss: 2.8624
Epoch 10: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3750 - loss: 2.8624 - val_accuracy: 0.5000 - val_loss: 2.8194
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4483 - loss: 2.8572
Epoch 11: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.4483 - loss: 2.8572 - val_accuracy: 0.4552 - val_loss: 2.8495
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:15 2s/step - accuracy: 0.3125 - loss: 2.8489
Epoch 12: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3125 - loss: 2.8489 - val_accuracy: 0.0000e+00 - val_loss: 2.8982
Epoch 13/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4727 - loss: 2.8454
Epoch 13: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 357s 2s/step - accuracy: 0.4727 - loss: 2.8454 - val_accuracy: 0.4783 - val_loss: 2.8412
Epoch 14/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.5000 - loss: 2.8418
Epoch 14: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5000 - loss: 2.8418 - val_accuracy: 0.0000e+00 - val_loss: 2.8537
Epoch 15/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4959 - loss: 2.8371
Epoch 15: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.4960 - loss: 2.8371 - val_accuracy: 0.4891 - val_loss: 2.8344
Epoch 16/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.6875 - loss: 2.8479
Epoch 16: val_accuracy improved from 0.50000 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 4s 14ms/step - accuracy: 0.6875 - loss: 2.8479 - val_accuracy: 1.0000 - val_loss: 2.7901
Epoch 17/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5275 - loss: 2.8315
Epoch 17: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 348s 2s/step - accuracy: 0.5275 - loss: 2.8315 - val_accuracy: 0.5204 - val_loss: 2.8290
Epoch 18/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.5625 - loss: 2.8137
Epoch 18: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5625 - loss: 2.8137 - val_accuracy: 0.0000e+00 - val_loss: 2.8259
Epoch 19/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5425 - loss: 2.8263
Epoch 19: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 356s 2s/step - accuracy: 0.5425 - loss: 2.8263 - val_accuracy: 0.5476 - val_loss: 2.8233
Epoch 20/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:18 2s/step - accuracy: 0.6250 - loss: 2.7994
Epoch 20: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6250 - loss: 2.7994 - val_accuracy: 0.5000 - val_loss: 2.7774
Epoch 21/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5683 - loss: 2.8195
Epoch 21: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.5683 - loss: 2.8195 - val_accuracy: 0.5516 - val_loss: 2.8182
Epoch 22/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.5625 - loss: 2.8063
Epoch 22: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5625 - loss: 2.8063 - val_accuracy: 1.0000 - val_loss: 2.7622
Epoch 23/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5696 - loss: 2.8150
Epoch 23: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.5695 - loss: 2.8150 - val_accuracy: 0.5516 - val_loss: 2.8129
Epoch 24/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:30 2s/step - accuracy: 0.6875 - loss: 2.7975
Epoch 24: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6875 - loss: 2.7975 - val_accuracy: 0.5000 - val_loss: 2.8173
Epoch 25/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5609 - loss: 2.8110
Epoch 25: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 360s 2s/step - accuracy: 0.5609 - loss: 2.8110 - val_accuracy: 0.5679 - val_loss: 2.8098
Epoch 26/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:15 2s/step - accuracy: 0.5000 - loss: 2.8230
Epoch 26: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5000 - loss: 2.8230 - val_accuracy: 0.0000e+00 - val_loss: 2.8437
Epoch 26: early stopping
Restoring model weights from the end of the best epoch: 16.
Tiempo transcurrido para el entrenamiento: 4630.410501003265 segundos
Uso de CPU durante el entrenamiento: 46.2%
Aumento en uso de memoria: -0.24663543701171875 GB
Resultados para lr=0.0001, l2=0.01, batch_size=16:
Tiempo: 4630.410501003265 segundos, CPU: 46.2%, Memoria: -0.24663543701171875 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_6"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_3 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.0962 - loss: 3.2078
Epoch 1: val_accuracy improved from -inf to 0.10598, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 346s 4s/step - accuracy: 0.0961 - loss: 3.2075 - val_accuracy: 0.1060 - val_loss: 3.1354
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:01 3s/step - accuracy: 0.1875 - loss: 3.1282
Epoch 2: val_accuracy did not improve from 0.10598
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.1875 - loss: 3.1282 - val_accuracy: 0.0000e+00 - val_loss: 3.1432
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.1299 - loss: 3.1149
Epoch 3: val_accuracy improved from 0.10598 to 0.13587, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 347s 4s/step - accuracy: 0.1300 - loss: 3.1147 - val_accuracy: 0.1359 - val_loss: 3.0595
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:59 3s/step - accuracy: 0.1250 - loss: 3.0502
Epoch 4: val_accuracy did not improve from 0.13587
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.1250 - loss: 3.0502 - val_accuracy: 0.0000e+00 - val_loss: 3.0644
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.1841 - loss: 3.0432
Epoch 5: val_accuracy improved from 0.13587 to 0.21196, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.1843 - loss: 3.0430 - val_accuracy: 0.2120 - val_loss: 3.0020
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:09 3s/step - accuracy: 0.2812 - loss: 2.9933
Epoch 6: val_accuracy did not improve from 0.21196
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.2812 - loss: 2.9933 - val_accuracy: 0.0000e+00 - val_loss: 3.0296
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.2350 - loss: 2.9884
Epoch 7: val_accuracy improved from 0.21196 to 0.25000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.2351 - loss: 2.9882 - val_accuracy: 0.2500 - val_loss: 2.9585
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.4375 - loss: 2.9432
Epoch 8: val_accuracy did not improve from 0.25000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.4375 - loss: 2.9432 - val_accuracy: 0.0000e+00 - val_loss: 3.0041
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.2725 - loss: 2.9482
Epoch 9: val_accuracy did not improve from 0.25000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.2724 - loss: 2.9481 - val_accuracy: 0.2459 - val_loss: 2.9271
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.2500 - loss: 2.9240
Epoch 10: val_accuracy did not improve from 0.25000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.2500 - loss: 2.9240 - val_accuracy: 0.0000e+00 - val_loss: 2.9464
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.3018 - loss: 2.9175
Epoch 11: val_accuracy improved from 0.25000 to 0.30163, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 346s 4s/step - accuracy: 0.3017 - loss: 2.9175 - val_accuracy: 0.3016 - val_loss: 2.9026
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.4688 - loss: 2.9051
Epoch 12: val_accuracy improved from 0.30163 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.4688 - loss: 2.9051 - val_accuracy: 1.0000 - val_loss: 2.8801
Epoch 13/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.3486 - loss: 2.8968
Epoch 13: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 352s 4s/step - accuracy: 0.3487 - loss: 2.8967 - val_accuracy: 0.3614 - val_loss: 2.8845
Epoch 14/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.3750 - loss: 2.8825
Epoch 14: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.3750 - loss: 2.8825 - val_accuracy: 0.5000 - val_loss: 2.8587
Epoch 15/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.3892 - loss: 2.8784
Epoch 15: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.3892 - loss: 2.8784 - val_accuracy: 0.3668 - val_loss: 2.8707
Epoch 16/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.4375 - loss: 2.8664
Epoch 16: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.4375 - loss: 2.8664 - val_accuracy: 1.0000 - val_loss: 2.8524
Epoch 17/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.3922 - loss: 2.8671
Epoch 17: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.3924 - loss: 2.8670 - val_accuracy: 0.3967 - val_loss: 2.8595
Epoch 18/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:34 3s/step - accuracy: 0.3750 - loss: 2.8577
Epoch 18: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.3750 - loss: 2.8577 - val_accuracy: 0.0000e+00 - val_loss: 2.8647
Epoch 19/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.4448 - loss: 2.8548
Epoch 19: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.4446 - loss: 2.8548 - val_accuracy: 0.4103 - val_loss: 2.8509
Epoch 20/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:13 3s/step - accuracy: 0.4375 - loss: 2.8489
Epoch 20: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.4375 - loss: 2.8489 - val_accuracy: 0.5000 - val_loss: 2.8608
Epoch 21/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.4260 - loss: 2.8478
Epoch 21: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.4262 - loss: 2.8478 - val_accuracy: 0.4552 - val_loss: 2.8449
Epoch 22/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.3750 - loss: 2.8640
Epoch 22: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.3750 - loss: 2.8640 - val_accuracy: 0.5000 - val_loss: 2.8430
Epoch 22: early stopping
Restoring model weights from the end of the best epoch: 12.
Tiempo transcurrido para el entrenamiento: 3835.0894253253937 segundos
Uso de CPU durante el entrenamiento: 55.2%
Aumento en uso de memoria: -0.19287872314453125 GB
Resultados para lr=0.0001, l2=0.01, batch_size=32:
Tiempo: 3835.0894253253937 segundos, CPU: 55.2%, Memoria: -0.19287872314453125 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_8"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_4 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.0637 - loss: 3.2282
Epoch 1: val_accuracy improved from -inf to 0.09375, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 356s 9s/step - accuracy: 0.0640 - loss: 3.2278 - val_accuracy: 0.0938 - val_loss: 3.1871
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.0469 - loss: 3.1898
Epoch 2: val_accuracy did not improve from 0.09375
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.0469 - loss: 3.1898 - val_accuracy: 0.0294 - val_loss: 3.1986
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.0889 - loss: 3.1741
Epoch 3: val_accuracy improved from 0.09375 to 0.13778, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.0893 - loss: 3.1738 - val_accuracy: 0.1378 - val_loss: 3.1378
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:32 7s/step - accuracy: 0.1562 - loss: 3.1376
Epoch 4: val_accuracy improved from 0.13778 to 0.23529, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 156ms/step - accuracy: 0.1562 - loss: 3.1376 - val_accuracy: 0.2353 - val_loss: 3.1327
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.1497 - loss: 3.1260
Epoch 5: val_accuracy did not improve from 0.23529
40/40 ━━━━━━━━━━━━━━━━━━━━ 343s 8s/step - accuracy: 0.1498 - loss: 3.1257 - val_accuracy: 0.1719 - val_loss: 3.0958
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.2812 - loss: 3.0909
Epoch 6: val_accuracy did not improve from 0.23529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.2812 - loss: 3.0909 - val_accuracy: 0.1176 - val_loss: 3.0941
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.1644 - loss: 3.0859
Epoch 7: val_accuracy did not improve from 0.23529
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.1649 - loss: 3.0856 - val_accuracy: 0.1932 - val_loss: 3.0592
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:15 7s/step - accuracy: 0.2344 - loss: 3.0572
Epoch 8: val_accuracy did not improve from 0.23529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.2344 - loss: 3.0572 - val_accuracy: 0.1765 - val_loss: 3.0660
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.2263 - loss: 3.0496
Epoch 9: val_accuracy did not improve from 0.23529
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.2263 - loss: 3.0494 - val_accuracy: 0.2287 - val_loss: 3.0282
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:03 6s/step - accuracy: 0.1406 - loss: 3.0338
Epoch 10: val_accuracy improved from 0.23529 to 0.26471, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 141ms/step - accuracy: 0.1406 - loss: 3.0338 - val_accuracy: 0.2647 - val_loss: 3.0259
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.2330 - loss: 3.0201
Epoch 11: val_accuracy did not improve from 0.26471
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.2332 - loss: 3.0199 - val_accuracy: 0.2585 - val_loss: 2.9999
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:15 7s/step - accuracy: 0.2656 - loss: 2.9968
Epoch 12: val_accuracy improved from 0.26471 to 0.29412, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 161ms/step - accuracy: 0.2656 - loss: 2.9968 - val_accuracy: 0.2941 - val_loss: 2.9980
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.2734 - loss: 2.9926
Epoch 13: val_accuracy did not improve from 0.29412
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.2735 - loss: 2.9924 - val_accuracy: 0.2926 - val_loss: 2.9758
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.2969 - loss: 2.9771
Epoch 14: val_accuracy did not improve from 0.29412
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.2969 - loss: 2.9771 - val_accuracy: 0.2059 - val_loss: 2.9774
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.2819 - loss: 2.9708
Epoch 15: val_accuracy did not improve from 0.29412
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.2820 - loss: 2.9707 - val_accuracy: 0.2827 - val_loss: 2.9567
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:17 7s/step - accuracy: 0.2969 - loss: 2.9522
Epoch 16: val_accuracy improved from 0.29412 to 0.38235, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 153ms/step - accuracy: 0.2969 - loss: 2.9522 - val_accuracy: 0.3824 - val_loss: 2.9540
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.3017 - loss: 2.9496
Epoch 17: val_accuracy did not improve from 0.38235
40/40 ━━━━━━━━━━━━━━━━━━━━ 344s 8s/step - accuracy: 0.3018 - loss: 2.9495 - val_accuracy: 0.3068 - val_loss: 2.9375
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.2500 - loss: 2.9362
Epoch 18: val_accuracy did not improve from 0.38235
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.2500 - loss: 2.9362 - val_accuracy: 0.2059 - val_loss: 2.9456
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.3082 - loss: 2.9340
Epoch 19: val_accuracy did not improve from 0.38235
40/40 ━━━━━━━━━━━━━━━━━━━━ 341s 8s/step - accuracy: 0.3085 - loss: 2.9339 - val_accuracy: 0.3509 - val_loss: 2.9224
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.3594 - loss: 2.9156
Epoch 20: val_accuracy did not improve from 0.38235
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.3594 - loss: 2.9156 - val_accuracy: 0.2941 - val_loss: 2.9260
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.3634 - loss: 2.9164
Epoch 21: val_accuracy did not improve from 0.38235
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.3635 - loss: 2.9163 - val_accuracy: 0.3509 - val_loss: 2.9096
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:06 6s/step - accuracy: 0.3281 - loss: 2.9138
Epoch 22: val_accuracy did not improve from 0.38235
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 94ms/step - accuracy: 0.3281 - loss: 2.9138 - val_accuracy: 0.3529 - val_loss: 2.9083
Epoch 23/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.3835 - loss: 2.9046
Epoch 23: val_accuracy did not improve from 0.38235
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.3835 - loss: 2.9046 - val_accuracy: 0.3693 - val_loss: 2.8968
Epoch 24/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:54 6s/step - accuracy: 0.2656 - loss: 2.9035
Epoch 24: val_accuracy improved from 0.38235 to 0.41176, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 143ms/step - accuracy: 0.2656 - loss: 2.9035 - val_accuracy: 0.4118 - val_loss: 2.8977
Epoch 25/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.3804 - loss: 2.8928
Epoch 25: val_accuracy did not improve from 0.41176
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.3805 - loss: 2.8928 - val_accuracy: 0.4006 - val_loss: 2.8877
Epoch 26/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.4688 - loss: 2.8915
Epoch 26: val_accuracy did not improve from 0.41176
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.4688 - loss: 2.8915 - val_accuracy: 0.3235 - val_loss: 2.8900
Epoch 27/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4233 - loss: 2.8832
Epoch 27: val_accuracy improved from 0.41176 to 0.41903, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.4235 - loss: 2.8831 - val_accuracy: 0.4190 - val_loss: 2.8796
Epoch 28/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:06 6s/step - accuracy: 0.4375 - loss: 2.8755
Epoch 28: val_accuracy did not improve from 0.41903
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.4375 - loss: 2.8755 - val_accuracy: 0.4118 - val_loss: 2.8802
Epoch 29/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4318 - loss: 2.8751
Epoch 29: val_accuracy improved from 0.41903 to 0.48864, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 347s 9s/step - accuracy: 0.4325 - loss: 2.8750 - val_accuracy: 0.4886 - val_loss: 2.8711
Epoch 30/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:48 7s/step - accuracy: 0.4219 - loss: 2.8805
Epoch 30: val_accuracy did not improve from 0.48864
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 93ms/step - accuracy: 0.4219 - loss: 2.8805 - val_accuracy: 0.4118 - val_loss: 2.8805
Epoch 31/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4551 - loss: 2.8678
Epoch 31: val_accuracy did not improve from 0.48864
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.4553 - loss: 2.8678 - val_accuracy: 0.4645 - val_loss: 2.8628
Epoch 32/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:06 6s/step - accuracy: 0.5156 - loss: 2.8593
Epoch 32: val_accuracy did not improve from 0.48864
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.5156 - loss: 2.8593 - val_accuracy: 0.4118 - val_loss: 2.8724
Epoch 33/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4645 - loss: 2.8606
Epoch 33: val_accuracy did not improve from 0.48864
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.4649 - loss: 2.8606 - val_accuracy: 0.4716 - val_loss: 2.8583
Epoch 34/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:13 6s/step - accuracy: 0.5938 - loss: 2.8492
Epoch 34: val_accuracy improved from 0.48864 to 0.52941, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 140ms/step - accuracy: 0.5938 - loss: 2.8492 - val_accuracy: 0.5294 - val_loss: 2.8563
Epoch 35/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5063 - loss: 2.8539
Epoch 35: val_accuracy did not improve from 0.52941
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.5062 - loss: 2.8539 - val_accuracy: 0.4773 - val_loss: 2.8527
Epoch 36/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.4062 - loss: 2.8470
Epoch 36: val_accuracy improved from 0.52941 to 0.55882, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 159ms/step - accuracy: 0.4062 - loss: 2.8470 - val_accuracy: 0.5588 - val_loss: 2.8492
Epoch 37/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4778 - loss: 2.8506
Epoch 37: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.4783 - loss: 2.8506 - val_accuracy: 0.4815 - val_loss: 2.8484
Epoch 38/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.5469 - loss: 2.8418
Epoch 38: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 96ms/step - accuracy: 0.5469 - loss: 2.8418 - val_accuracy: 0.5294 - val_loss: 2.8515
Epoch 39/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5024 - loss: 2.8463
Epoch 39: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.5027 - loss: 2.8463 - val_accuracy: 0.5312 - val_loss: 2.8417
Epoch 40/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.4062 - loss: 2.8479
Epoch 40: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 84ms/step - accuracy: 0.4062 - loss: 2.8479 - val_accuracy: 0.5588 - val_loss: 2.8566
Epoch 41/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5246 - loss: 2.8414
Epoch 41: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.5244 - loss: 2.8413 - val_accuracy: 0.5014 - val_loss: 2.8398
Epoch 42/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:15 7s/step - accuracy: 0.5000 - loss: 2.8380
Epoch 42: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 99ms/step - accuracy: 0.5000 - loss: 2.8380 - val_accuracy: 0.5294 - val_loss: 2.8391
Epoch 43/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5340 - loss: 2.8378
Epoch 43: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 345s 9s/step - accuracy: 0.5337 - loss: 2.8378 - val_accuracy: 0.5028 - val_loss: 2.8368
Epoch 44/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.5781 - loss: 2.8277
Epoch 44: val_accuracy improved from 0.55882 to 0.58824, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 138ms/step - accuracy: 0.5781 - loss: 2.8277 - val_accuracy: 0.5882 - val_loss: 2.8429
Epoch 45/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5235 - loss: 2.8362
Epoch 45: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.5236 - loss: 2.8361 - val_accuracy: 0.5241 - val_loss: 2.8325
Epoch 46/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.5000 - loss: 2.8399
Epoch 46: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.5000 - loss: 2.8399 - val_accuracy: 0.5588 - val_loss: 2.8369
Epoch 47/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5276 - loss: 2.8321
Epoch 47: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.5277 - loss: 2.8321 - val_accuracy: 0.5142 - val_loss: 2.8300
Epoch 48/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.5000 - loss: 2.8242
Epoch 48: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 133ms/step - accuracy: 0.5000 - loss: 2.8242 - val_accuracy: 0.5294 - val_loss: 2.8356
Epoch 49/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5265 - loss: 2.8284
Epoch 49: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.5266 - loss: 2.8284 - val_accuracy: 0.5398 - val_loss: 2.8271
Epoch 50/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.5000 - loss: 2.8265
Epoch 50: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.5000 - loss: 2.8265 - val_accuracy: 0.5294 - val_loss: 2.8228
Restoring model weights from the end of the best epoch: 44.
Tiempo transcurrido para el entrenamiento: 8753.506147384644 segundos
Uso de CPU durante el entrenamiento: 43.8%
Aumento en uso de memoria: 0.314056396484375 GB
Resultados para lr=0.0001, l2=0.01, batch_size=64:
Tiempo: 8753.506147384644 segundos, CPU: 43.8%, Memoria: 0.314056396484375 GB
Precisión de validación: 0.5882353186607361
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_10"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_5 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.1113 - loss: 3.1941
Epoch 1: val_accuracy improved from -inf to 0.20516, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 357s 2s/step - accuracy: 0.1115 - loss: 3.1938 - val_accuracy: 0.2052 - val_loss: 3.0679
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 6:39 2s/step - accuracy: 0.2500 - loss: 3.0733
Epoch 2: val_accuracy did not improve from 0.20516
163/163 ━━━━━━━━━━━━━━━━━━━━ 3s 3ms/step - accuracy: 0.2500 - loss: 3.0733 - val_accuracy: 0.0000e+00 - val_loss: 3.0653
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.2256 - loss: 3.0380
Epoch 3: val_accuracy improved from 0.20516 to 0.28261, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 356s 2s/step - accuracy: 0.2257 - loss: 3.0378 - val_accuracy: 0.2826 - val_loss: 2.9675
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 0.1875 - loss: 2.9712
Epoch 4: val_accuracy did not improve from 0.28261
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.1875 - loss: 2.9712 - val_accuracy: 0.0000e+00 - val_loss: 2.9652
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.2843 - loss: 2.9506
Epoch 5: val_accuracy improved from 0.28261 to 0.33967, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 352s 2s/step - accuracy: 0.2843 - loss: 2.9505 - val_accuracy: 0.3397 - val_loss: 2.9104
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:05 2s/step - accuracy: 0.4375 - loss: 2.9084
Epoch 6: val_accuracy improved from 0.33967 to 0.50000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 4s 13ms/step - accuracy: 0.4375 - loss: 2.9084 - val_accuracy: 0.5000 - val_loss: 2.8842
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.3414 - loss: 2.9013
Epoch 7: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 357s 2s/step - accuracy: 0.3415 - loss: 2.9012 - val_accuracy: 0.4117 - val_loss: 2.8783
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:23 2s/step - accuracy: 0.5000 - loss: 2.8763
Epoch 8: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5000 - loss: 2.8763 - val_accuracy: 0.0000e+00 - val_loss: 2.9139
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.3987 - loss: 2.8723
Epoch 9: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.3988 - loss: 2.8723 - val_accuracy: 0.4266 - val_loss: 2.8597
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.2500 - loss: 2.8586
Epoch 10: val_accuracy improved from 0.50000 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 4s 14ms/step - accuracy: 0.2500 - loss: 2.8586 - val_accuracy: 1.0000 - val_loss: 2.8077
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4308 - loss: 2.8545
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.4308 - loss: 2.8545 - val_accuracy: 0.4429 - val_loss: 2.8485
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:20 2s/step - accuracy: 0.4375 - loss: 2.8701
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.4375 - loss: 2.8701 - val_accuracy: 1.0000 - val_loss: 2.8724
Epoch 13/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4579 - loss: 2.8433
Epoch 13: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 353s 2s/step - accuracy: 0.4579 - loss: 2.8433 - val_accuracy: 0.4484 - val_loss: 2.8391
Epoch 14/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.2500 - loss: 2.8536
Epoch 14: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.2500 - loss: 2.8536 - val_accuracy: 0.5000 - val_loss: 2.8715
Epoch 15/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4573 - loss: 2.8368
Epoch 15: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.4573 - loss: 2.8368 - val_accuracy: 0.4769 - val_loss: 2.8322
Epoch 16/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 0.5000 - loss: 2.8358
Epoch 16: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5000 - loss: 2.8358 - val_accuracy: 0.5000 - val_loss: 2.8970
Epoch 17/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4739 - loss: 2.8314
Epoch 17: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.4739 - loss: 2.8314 - val_accuracy: 0.4837 - val_loss: 2.8269
Epoch 18/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.3750 - loss: 2.8284
Epoch 18: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3750 - loss: 2.8284 - val_accuracy: 0.5000 - val_loss: 2.8102
Epoch 19/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4811 - loss: 2.8263
Epoch 19: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 357s 2s/step - accuracy: 0.4810 - loss: 2.8263 - val_accuracy: 0.4769 - val_loss: 2.8212
Epoch 20/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:15 2s/step - accuracy: 0.3125 - loss: 2.8236
Epoch 20: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3125 - loss: 2.8236 - val_accuracy: 0.0000e+00 - val_loss: 2.7921
Epoch 20: early stopping
Restoring model weights from the end of the best epoch: 10.
Tiempo transcurrido para el entrenamiento: 3554.6360399723053 segundos
Uso de CPU durante el entrenamiento: 56.0%
Aumento en uso de memoria: -1.0743408203125 GB
Resultados para lr=0.0001, l2=0.05, batch_size=16:
Tiempo: 3554.6360399723053 segundos, CPU: 56.0%, Memoria: -1.0743408203125 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_12"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_6 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.0846 - loss: 3.2216
Epoch 1: val_accuracy improved from -inf to 0.12228, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 348s 4s/step - accuracy: 0.0847 - loss: 3.2213 - val_accuracy: 0.1223 - val_loss: 3.1471
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:04 3s/step - accuracy: 0.0625 - loss: 3.1451
Epoch 2: val_accuracy improved from 0.12228 to 0.50000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 5s 27ms/step - accuracy: 0.0625 - loss: 3.1451 - val_accuracy: 0.5000 - val_loss: 3.1167
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.1387 - loss: 3.1243
Epoch 3: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 347s 4s/step - accuracy: 0.1389 - loss: 3.1241 - val_accuracy: 0.1753 - val_loss: 3.0696
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:14 3s/step - accuracy: 0.0938 - loss: 3.0729
Epoch 4: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.0938 - loss: 3.0729 - val_accuracy: 0.0000e+00 - val_loss: 3.0729
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.1962 - loss: 3.0523
Epoch 5: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.1964 - loss: 3.0521 - val_accuracy: 0.2242 - val_loss: 3.0106
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:04 3s/step - accuracy: 0.1250 - loss: 3.0209
Epoch 6: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.1250 - loss: 3.0209 - val_accuracy: 0.5000 - val_loss: 2.9899
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.2342 - loss: 2.9975
Epoch 7: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.2344 - loss: 2.9974 - val_accuracy: 0.2527 - val_loss: 2.9667
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:08 3s/step - accuracy: 0.3438 - loss: 2.9624
Epoch 8: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.3438 - loss: 2.9624 - val_accuracy: 0.5000 - val_loss: 2.9529
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.2962 - loss: 2.9548
Epoch 9: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.2961 - loss: 2.9547 - val_accuracy: 0.2867 - val_loss: 2.9333
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:17 3s/step - accuracy: 0.3438 - loss: 2.9273
Epoch 10: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.3438 - loss: 2.9273 - val_accuracy: 0.0000e+00 - val_loss: 2.9402
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.3170 - loss: 2.9251
Epoch 11: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.3171 - loss: 2.9250 - val_accuracy: 0.3356 - val_loss: 2.9085
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.4688 - loss: 2.9038
Epoch 12: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.4688 - loss: 2.9038 - val_accuracy: 0.5000 - val_loss: 2.9244
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2090.1293630599976 segundos
Uso de CPU durante el entrenamiento: 35.7%
Aumento en uso de memoria: -0.5012664794921875 GB
Resultados para lr=0.0001, l2=0.05, batch_size=32:
Tiempo: 2090.1293630599976 segundos, CPU: 35.7%, Memoria: -0.5012664794921875 GB
Precisión de validación: 0.5
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_14"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_7 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.0560 - loss: 3.2283
Epoch 1: val_accuracy improved from -inf to 0.08665, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 362s 9s/step - accuracy: 0.0561 - loss: 3.2279 - val_accuracy: 0.0866 - val_loss: 3.1875
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:07 6s/step - accuracy: 0.0781 - loss: 3.1928
Epoch 2: val_accuracy improved from 0.08665 to 0.11765, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 182ms/step - accuracy: 0.0781 - loss: 3.1928 - val_accuracy: 0.1176 - val_loss: 3.1972
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.0993 - loss: 3.1750
Epoch 3: val_accuracy improved from 0.11765 to 0.14062, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 341s 8s/step - accuracy: 0.0997 - loss: 3.1747 - val_accuracy: 0.1406 - val_loss: 3.1395
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:14 7s/step - accuracy: 0.0781 - loss: 3.1400
Epoch 4: val_accuracy improved from 0.14062 to 0.17647, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 154ms/step - accuracy: 0.0781 - loss: 3.1400 - val_accuracy: 0.1765 - val_loss: 3.1351
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.1304 - loss: 3.1284
Epoch 5: val_accuracy did not improve from 0.17647
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.1306 - loss: 3.1282 - val_accuracy: 0.1378 - val_loss: 3.0979
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:05 6s/step - accuracy: 0.1562 - loss: 3.0968
Epoch 6: val_accuracy did not improve from 0.17647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 90ms/step - accuracy: 0.1562 - loss: 3.0968 - val_accuracy: 0.1471 - val_loss: 3.0951
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.1588 - loss: 3.0876
Epoch 7: val_accuracy improved from 0.17647 to 0.21449, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.1593 - loss: 3.0873 - val_accuracy: 0.2145 - val_loss: 3.0602
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:03 6s/step - accuracy: 0.2031 - loss: 3.0544
Epoch 8: val_accuracy did not improve from 0.21449
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 84ms/step - accuracy: 0.2031 - loss: 3.0544 - val_accuracy: 0.1765 - val_loss: 3.0659
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.2059 - loss: 3.0523
Epoch 9: val_accuracy improved from 0.21449 to 0.25710, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.2061 - loss: 3.0521 - val_accuracy: 0.2571 - val_loss: 3.0281
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.2812 - loss: 3.0274
Epoch 10: val_accuracy did not improve from 0.25710
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 84ms/step - accuracy: 0.2812 - loss: 3.0274 - val_accuracy: 0.2353 - val_loss: 3.0325
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.2654 - loss: 3.0205
Epoch 11: val_accuracy improved from 0.25710 to 0.28693, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 341s 8s/step - accuracy: 0.2656 - loss: 3.0204 - val_accuracy: 0.2869 - val_loss: 3.0010
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:42 7s/step - accuracy: 0.2500 - loss: 3.0018
Epoch 12: val_accuracy did not improve from 0.28693
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 97ms/step - accuracy: 0.2500 - loss: 3.0018 - val_accuracy: 0.2647 - val_loss: 3.0013
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.2995 - loss: 2.9953
Epoch 13: val_accuracy improved from 0.28693 to 0.30824, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.2994 - loss: 2.9952 - val_accuracy: 0.3082 - val_loss: 2.9767
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.2812 - loss: 2.9798
Epoch 14: val_accuracy did not improve from 0.30824
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.2812 - loss: 2.9798 - val_accuracy: 0.2353 - val_loss: 2.9869
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.3038 - loss: 2.9719
Epoch 15: val_accuracy improved from 0.30824 to 0.31676, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 347s 9s/step - accuracy: 0.3039 - loss: 2.9718 - val_accuracy: 0.3168 - val_loss: 2.9564
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:05 6s/step - accuracy: 0.3750 - loss: 2.9548
Epoch 16: val_accuracy did not improve from 0.31676
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 93ms/step - accuracy: 0.3750 - loss: 2.9548 - val_accuracy: 0.2647 - val_loss: 2.9630
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.3335 - loss: 2.9509
Epoch 17: val_accuracy improved from 0.31676 to 0.33381, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.3332 - loss: 2.9508 - val_accuracy: 0.3338 - val_loss: 2.9390
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.3125 - loss: 2.9390
Epoch 18: val_accuracy improved from 0.33381 to 0.35294, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 143ms/step - accuracy: 0.3125 - loss: 2.9390 - val_accuracy: 0.3529 - val_loss: 2.9348
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.3304 - loss: 2.9348
Epoch 19: val_accuracy improved from 0.35294 to 0.37358, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.3307 - loss: 2.9347 - val_accuracy: 0.3736 - val_loss: 2.9227
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.3594 - loss: 2.9195
Epoch 20: val_accuracy did not improve from 0.37358
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.3594 - loss: 2.9195 - val_accuracy: 0.2647 - val_loss: 2.9281
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.3927 - loss: 2.9194
Epoch 21: val_accuracy improved from 0.37358 to 0.38920, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 342s 8s/step - accuracy: 0.3924 - loss: 2.9193 - val_accuracy: 0.3892 - val_loss: 2.9089
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:19 7s/step - accuracy: 0.3594 - loss: 2.9123
Epoch 22: val_accuracy did not improve from 0.38920
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.3594 - loss: 2.9123 - val_accuracy: 0.3529 - val_loss: 2.9151
Epoch 23/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.3828 - loss: 2.9063
Epoch 23: val_accuracy improved from 0.38920 to 0.41051, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.3834 - loss: 2.9063 - val_accuracy: 0.4105 - val_loss: 2.8982
Epoch 24/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:17 7s/step - accuracy: 0.3906 - loss: 2.8971
Epoch 24: val_accuracy improved from 0.41051 to 0.44118, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 145ms/step - accuracy: 0.3906 - loss: 2.8971 - val_accuracy: 0.4412 - val_loss: 2.8982
Epoch 25/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4000 - loss: 2.8953
Epoch 25: val_accuracy did not improve from 0.44118
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.4002 - loss: 2.8952 - val_accuracy: 0.4105 - val_loss: 2.8882
Epoch 26/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:17 7s/step - accuracy: 0.2969 - loss: 2.8918
Epoch 26: val_accuracy improved from 0.44118 to 0.50000, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 151ms/step - accuracy: 0.2969 - loss: 2.8918 - val_accuracy: 0.5000 - val_loss: 2.8780
Epoch 27/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4405 - loss: 2.8842
Epoch 27: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 345s 9s/step - accuracy: 0.4405 - loss: 2.8842 - val_accuracy: 0.4503 - val_loss: 2.8787
Epoch 28/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.3750 - loss: 2.8840
Epoch 28: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.3750 - loss: 2.8840 - val_accuracy: 0.3824 - val_loss: 2.8780
Epoch 29/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4500 - loss: 2.8754
Epoch 29: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 335s 8s/step - accuracy: 0.4500 - loss: 2.8754 - val_accuracy: 0.4531 - val_loss: 2.8710
Epoch 30/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:05 6s/step - accuracy: 0.4062 - loss: 2.8761
Epoch 30: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 95ms/step - accuracy: 0.4062 - loss: 2.8761 - val_accuracy: 0.5000 - val_loss: 2.8717
Epoch 31/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4720 - loss: 2.8684
Epoch 31: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.4716 - loss: 2.8683 - val_accuracy: 0.4659 - val_loss: 2.8644
Epoch 32/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.5469 - loss: 2.8574
Epoch 32: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 91ms/step - accuracy: 0.5469 - loss: 2.8574 - val_accuracy: 0.4706 - val_loss: 2.8643
Epoch 33/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4668 - loss: 2.8623
Epoch 33: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.4669 - loss: 2.8623 - val_accuracy: 0.4986 - val_loss: 2.8580
Epoch 34/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.4844 - loss: 2.8568
Epoch 34: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.4844 - loss: 2.8568 - val_accuracy: 0.4412 - val_loss: 2.8559
Epoch 35/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4727 - loss: 2.8555
Epoch 35: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.4727 - loss: 2.8555 - val_accuracy: 0.4787 - val_loss: 2.8529
Epoch 36/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:23 7s/step - accuracy: 0.4688 - loss: 2.8436
Epoch 36: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 99ms/step - accuracy: 0.4688 - loss: 2.8436 - val_accuracy: 0.3824 - val_loss: 2.8607
Epoch 36: early stopping
Restoring model weights from the end of the best epoch: 26.
Tiempo transcurrido para el entrenamiento: 6330.446506977081 segundos
Uso de CPU durante el entrenamiento: 40.89999999999999%
Aumento en uso de memoria: 0.4019775390625 GB
Resultados para lr=0.0001, l2=0.05, batch_size=64:
Tiempo: 6330.446506977081 segundos, CPU: 40.89999999999999%, Memoria: 0.4019775390625 GB
Precisión de validación: 0.5
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_16"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_8 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.0853 - loss: 3.1943
Epoch 1: val_accuracy improved from -inf to 0.16984, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 355s 2s/step - accuracy: 0.0855 - loss: 3.1940 - val_accuracy: 0.1698 - val_loss: 3.0692
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:35 2s/step - accuracy: 0.2500 - loss: 3.0561
Epoch 2: val_accuracy improved from 0.16984 to 0.50000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 5s 22ms/step - accuracy: 0.2500 - loss: 3.0561 - val_accuracy: 0.5000 - val_loss: 3.0764
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.2061 - loss: 3.0397
Epoch 3: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 359s 2s/step - accuracy: 0.2062 - loss: 3.0395 - val_accuracy: 0.2473 - val_loss: 2.9681
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:48 2s/step - accuracy: 0.1875 - loss: 2.9743
Epoch 4: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.1875 - loss: 2.9743 - val_accuracy: 0.0000e+00 - val_loss: 2.9952
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.2686 - loss: 2.9525
Epoch 5: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.2686 - loss: 2.9524 - val_accuracy: 0.2772 - val_loss: 2.9127
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.3750 - loss: 2.9119
Epoch 6: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3750 - loss: 2.9119 - val_accuracy: 0.5000 - val_loss: 2.9348
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.3174 - loss: 2.9038
Epoch 7: val_accuracy did not improve from 0.50000
163/163 ━━━━━━━━━━━━━━━━━━━━ 352s 2s/step - accuracy: 0.3174 - loss: 2.9038 - val_accuracy: 0.3329 - val_loss: 2.8813
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:23 2s/step - accuracy: 0.5625 - loss: 2.8722
Epoch 8: val_accuracy improved from 0.50000 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 4s 13ms/step - accuracy: 0.5625 - loss: 2.8722 - val_accuracy: 1.0000 - val_loss: 2.8368
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.3585 - loss: 2.8760
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.3586 - loss: 2.8760 - val_accuracy: 0.3804 - val_loss: 2.8621
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:25 2s/step - accuracy: 0.3750 - loss: 2.8686
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3750 - loss: 2.8686 - val_accuracy: 1.0000 - val_loss: 2.8165
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.3938 - loss: 2.8580
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.3939 - loss: 2.8580 - val_accuracy: 0.4348 - val_loss: 2.8503
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.3750 - loss: 2.8423
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3750 - loss: 2.8423 - val_accuracy: 0.0000e+00 - val_loss: 2.8909
Epoch 13/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4700 - loss: 2.8483
Epoch 13: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 348s 2s/step - accuracy: 0.4700 - loss: 2.8483 - val_accuracy: 0.4701 - val_loss: 2.8421
Epoch 14/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:15 2s/step - accuracy: 0.5000 - loss: 2.8387
Epoch 14: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5000 - loss: 2.8387 - val_accuracy: 0.5000 - val_loss: 2.8016
Epoch 15/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4747 - loss: 2.8408
Epoch 15: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.4748 - loss: 2.8408 - val_accuracy: 0.4783 - val_loss: 2.8360
Epoch 16/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.3750 - loss: 2.8509
Epoch 16: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3750 - loss: 2.8509 - val_accuracy: 0.5000 - val_loss: 2.8490
Epoch 17/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4948 - loss: 2.8336
Epoch 17: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 361s 2s/step - accuracy: 0.4948 - loss: 2.8336 - val_accuracy: 0.4973 - val_loss: 2.8295
Epoch 18/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.3750 - loss: 2.8351
Epoch 18: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3750 - loss: 2.8351 - val_accuracy: 0.0000e+00 - val_loss: 2.8519
Epoch 18: early stopping
Restoring model weights from the end of the best epoch: 8.
Tiempo transcurrido para el entrenamiento: 3196.9851167201996 segundos
Uso de CPU durante el entrenamiento: 41.3%
Aumento en uso de memoria: -0.6323471069335938 GB
Resultados para lr=0.0001, l2=0.1, batch_size=16:
Tiempo: 3196.9851167201996 segundos, CPU: 41.3%, Memoria: -0.6323471069335938 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_18"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_9 (InputLayer)      │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.0545 - loss: 3.2192
Epoch 1: val_accuracy improved from -inf to 0.07201, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 349s 4s/step - accuracy: 0.0545 - loss: 3.2189 - val_accuracy: 0.0720 - val_loss: 3.1452
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:08 3s/step - accuracy: 0.1250 - loss: 3.1491
Epoch 2: val_accuracy improved from 0.07201 to 0.50000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 6s 31ms/step - accuracy: 0.1250 - loss: 3.1491 - val_accuracy: 0.5000 - val_loss: 3.1301
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.0782 - loss: 3.1235
Epoch 3: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.0785 - loss: 3.1233 - val_accuracy: 0.1168 - val_loss: 3.0672
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.0938 - loss: 3.0692
Epoch 4: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.0938 - loss: 3.0692 - val_accuracy: 0.0000e+00 - val_loss: 3.0670
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.1578 - loss: 3.0509
Epoch 5: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.1580 - loss: 3.0507 - val_accuracy: 0.2065 - val_loss: 3.0084
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:18 3s/step - accuracy: 0.1875 - loss: 3.0112
Epoch 6: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.1875 - loss: 3.0112 - val_accuracy: 0.0000e+00 - val_loss: 3.0557
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.2360 - loss: 2.9954
Epoch 7: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 341s 4s/step - accuracy: 0.2361 - loss: 2.9953 - val_accuracy: 0.2582 - val_loss: 2.9648
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:02 3s/step - accuracy: 0.3750 - loss: 2.9655
Epoch 8: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.3750 - loss: 2.9655 - val_accuracy: 0.0000e+00 - val_loss: 2.9459
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.2611 - loss: 2.9551
Epoch 9: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.2614 - loss: 2.9550 - val_accuracy: 0.2880 - val_loss: 2.9307
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:09 3s/step - accuracy: 0.3125 - loss: 2.9365
Epoch 10: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.3125 - loss: 2.9365 - val_accuracy: 0.0000e+00 - val_loss: 2.8752
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.3115 - loss: 2.9233
Epoch 11: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 350s 4s/step - accuracy: 0.3115 - loss: 2.9232 - val_accuracy: 0.3152 - val_loss: 2.9061
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:04 3s/step - accuracy: 0.4062 - loss: 2.9018
Epoch 12: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.4062 - loss: 2.9018 - val_accuracy: 0.5000 - val_loss: 2.8846
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2093.3566970825195 segundos
Uso de CPU durante el entrenamiento: 41.4%
Aumento en uso de memoria: -0.2600898742675781 GB
Resultados para lr=0.0001, l2=0.1, batch_size=32:
Tiempo: 2093.3566970825195 segundos, CPU: 41.4%, Memoria: -0.2600898742675781 GB
Precisión de validación: 0.5
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_20"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_10 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.0383 - loss: 3.2350
Epoch 1: val_accuracy improved from -inf to 0.05682, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 349s 8s/step - accuracy: 0.0384 - loss: 3.2346 - val_accuracy: 0.0568 - val_loss: 3.1941
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.0469 - loss: 3.1963
Epoch 2: val_accuracy did not improve from 0.05682
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 91ms/step - accuracy: 0.0469 - loss: 3.1963 - val_accuracy: 0.0294 - val_loss: 3.1926
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.0603 - loss: 3.1807
Epoch 3: val_accuracy improved from 0.05682 to 0.08381, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.0605 - loss: 3.1804 - val_accuracy: 0.0838 - val_loss: 3.1449
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.0781 - loss: 3.1474
Epoch 4: val_accuracy did not improve from 0.08381
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.0781 - loss: 3.1474 - val_accuracy: 0.0588 - val_loss: 3.1452
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.0955 - loss: 3.1337
Epoch 5: val_accuracy improved from 0.08381 to 0.13778, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 341s 8s/step - accuracy: 0.0959 - loss: 3.1335 - val_accuracy: 0.1378 - val_loss: 3.1032
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.1094 - loss: 3.1056
Epoch 6: val_accuracy improved from 0.13778 to 0.29412, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 142ms/step - accuracy: 0.1094 - loss: 3.1056 - val_accuracy: 0.2941 - val_loss: 3.1019
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.1770 - loss: 3.0917
Epoch 7: val_accuracy did not improve from 0.29412
40/40 ━━━━━━━━━━━━━━━━━━━━ 342s 8s/step - accuracy: 0.1767 - loss: 3.0915 - val_accuracy: 0.1705 - val_loss: 3.0664
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.1875 - loss: 3.0669
Epoch 8: val_accuracy did not improve from 0.29412
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 87ms/step - accuracy: 0.1875 - loss: 3.0669 - val_accuracy: 0.1471 - val_loss: 3.0635
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.1764 - loss: 3.0575
Epoch 9: val_accuracy did not improve from 0.29412
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.1769 - loss: 3.0573 - val_accuracy: 0.2145 - val_loss: 3.0345
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.2188 - loss: 3.0314
Epoch 10: val_accuracy did not improve from 0.29412
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 86ms/step - accuracy: 0.2188 - loss: 3.0314 - val_accuracy: 0.2647 - val_loss: 3.0337
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.2328 - loss: 3.0264
Epoch 11: val_accuracy did not improve from 0.29412
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.2330 - loss: 3.0262 - val_accuracy: 0.2429 - val_loss: 3.0069
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.1875 - loss: 3.0114
Epoch 12: val_accuracy did not improve from 0.29412
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 91ms/step - accuracy: 0.1875 - loss: 3.0114 - val_accuracy: 0.1765 - val_loss: 3.0064
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.2490 - loss: 2.9998
Epoch 13: val_accuracy did not improve from 0.29412
40/40 ━━━━━━━━━━━━━━━━━━━━ 344s 9s/step - accuracy: 0.2492 - loss: 2.9997 - val_accuracy: 0.2557 - val_loss: 2.9832
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.1719 - loss: 2.9882
Epoch 14: val_accuracy improved from 0.29412 to 0.35294, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 150ms/step - accuracy: 0.1719 - loss: 2.9882 - val_accuracy: 0.3529 - val_loss: 2.9832
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.2663 - loss: 2.9766
Epoch 15: val_accuracy did not improve from 0.35294
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.2667 - loss: 2.9765 - val_accuracy: 0.2628 - val_loss: 2.9619
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.2812 - loss: 2.9575
Epoch 16: val_accuracy did not improve from 0.35294
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.2812 - loss: 2.9575 - val_accuracy: 0.3235 - val_loss: 2.9610
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.2868 - loss: 2.9573
Epoch 17: val_accuracy did not improve from 0.35294
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.2872 - loss: 2.9572 - val_accuracy: 0.3168 - val_loss: 2.9440
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.4062 - loss: 2.9373
Epoch 18: val_accuracy did not improve from 0.35294
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.4062 - loss: 2.9373 - val_accuracy: 0.2941 - val_loss: 2.9432
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.3289 - loss: 2.9390
Epoch 19: val_accuracy did not improve from 0.35294
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.3291 - loss: 2.9389 - val_accuracy: 0.3395 - val_loss: 2.9284
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:48 7s/step - accuracy: 0.3125 - loss: 2.9247
Epoch 20: val_accuracy did not improve from 0.35294
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 86ms/step - accuracy: 0.3125 - loss: 2.9247 - val_accuracy: 0.1765 - val_loss: 2.9330
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.3539 - loss: 2.9242
Epoch 21: val_accuracy improved from 0.35294 to 0.36080, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 347s 9s/step - accuracy: 0.3540 - loss: 2.9241 - val_accuracy: 0.3608 - val_loss: 2.9149
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:55 8s/step - accuracy: 0.3438 - loss: 2.9140
Epoch 22: val_accuracy improved from 0.36080 to 0.44118, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 16s 218ms/step - accuracy: 0.3438 - loss: 2.9140 - val_accuracy: 0.4412 - val_loss: 2.9103
Epoch 23/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.3800 - loss: 2.9108
Epoch 23: val_accuracy did not improve from 0.44118
40/40 ━━━━━━━━━━━━━━━━━━━━ 345s 8s/step - accuracy: 0.3801 - loss: 2.9107 - val_accuracy: 0.3764 - val_loss: 2.9036
Epoch 24/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:08 6s/step - accuracy: 0.3438 - loss: 2.9003
Epoch 24: val_accuracy improved from 0.44118 to 0.47059, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 154ms/step - accuracy: 0.3438 - loss: 2.9003 - val_accuracy: 0.4706 - val_loss: 2.9024
Epoch 25/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.3783 - loss: 2.8986
Epoch 25: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 341s 8s/step - accuracy: 0.3786 - loss: 2.8986 - val_accuracy: 0.3949 - val_loss: 2.8936
Epoch 26/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:12 6s/step - accuracy: 0.3906 - loss: 2.8955
Epoch 26: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.3906 - loss: 2.8955 - val_accuracy: 0.4412 - val_loss: 2.8852
Epoch 27/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.3969 - loss: 2.8893
Epoch 27: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 348s 9s/step - accuracy: 0.3974 - loss: 2.8892 - val_accuracy: 0.4176 - val_loss: 2.8838
Epoch 28/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.3438 - loss: 2.8876
Epoch 28: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.3438 - loss: 2.8876 - val_accuracy: 0.4706 - val_loss: 2.8887
Epoch 29/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4356 - loss: 2.8802
Epoch 29: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.4354 - loss: 2.8802 - val_accuracy: 0.4219 - val_loss: 2.8756
Epoch 30/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:12 6s/step - accuracy: 0.3906 - loss: 2.8805
Epoch 30: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.3906 - loss: 2.8805 - val_accuracy: 0.4412 - val_loss: 2.8788
Epoch 31/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4370 - loss: 2.8744
Epoch 31: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 335s 8s/step - accuracy: 0.4370 - loss: 2.8744 - val_accuracy: 0.4176 - val_loss: 2.8696
Epoch 32/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.5156 - loss: 2.8642
Epoch 32: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.5156 - loss: 2.8642 - val_accuracy: 0.4706 - val_loss: 2.8600
Epoch 33/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4536 - loss: 2.8662
Epoch 33: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.4534 - loss: 2.8662 - val_accuracy: 0.4460 - val_loss: 2.8640
Epoch 34/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.5469 - loss: 2.8608
Epoch 34: val_accuracy improved from 0.47059 to 0.52941, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 139ms/step - accuracy: 0.5469 - loss: 2.8608 - val_accuracy: 0.5294 - val_loss: 2.8583
Epoch 35/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4732 - loss: 2.8603
Epoch 35: val_accuracy did not improve from 0.52941
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.4729 - loss: 2.8603 - val_accuracy: 0.4474 - val_loss: 2.8579
Epoch 36/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:10 6s/step - accuracy: 0.4219 - loss: 2.8595
Epoch 36: val_accuracy did not improve from 0.52941
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.4219 - loss: 2.8595 - val_accuracy: 0.4706 - val_loss: 2.8658
Epoch 37/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4609 - loss: 2.8542
Epoch 37: val_accuracy did not improve from 0.52941
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.4609 - loss: 2.8542 - val_accuracy: 0.4418 - val_loss: 2.8532
Epoch 38/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.5156 - loss: 2.8548
Epoch 38: val_accuracy did not improve from 0.52941
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 94ms/step - accuracy: 0.5156 - loss: 2.8548 - val_accuracy: 0.4706 - val_loss: 2.8553
Epoch 39/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4852 - loss: 2.8504
Epoch 39: val_accuracy did not improve from 0.52941
40/40 ━━━━━━━━━━━━━━━━━━━━ 343s 8s/step - accuracy: 0.4849 - loss: 2.8503 - val_accuracy: 0.4602 - val_loss: 2.8487
Epoch 40/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:55 6s/step - accuracy: 0.3750 - loss: 2.8548
Epoch 40: val_accuracy did not improve from 0.52941
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 95ms/step - accuracy: 0.3750 - loss: 2.8548 - val_accuracy: 0.3824 - val_loss: 2.8422
Epoch 41/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4663 - loss: 2.8471
Epoch 41: val_accuracy did not improve from 0.52941
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.4667 - loss: 2.8471 - val_accuracy: 0.4702 - val_loss: 2.8452
Epoch 42/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.5625 - loss: 2.8385
Epoch 42: val_accuracy improved from 0.52941 to 0.61765, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 152ms/step - accuracy: 0.5625 - loss: 2.8385 - val_accuracy: 0.6176 - val_loss: 2.8378
Epoch 43/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4842 - loss: 2.8435
Epoch 43: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.4843 - loss: 2.8435 - val_accuracy: 0.4716 - val_loss: 2.8423
Epoch 44/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.4844 - loss: 2.8379
Epoch 44: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 106ms/step - accuracy: 0.4844 - loss: 2.8379 - val_accuracy: 0.4412 - val_loss: 2.8418
Epoch 45/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4913 - loss: 2.8397
Epoch 45: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 335s 8s/step - accuracy: 0.4913 - loss: 2.8396 - val_accuracy: 0.4872 - val_loss: 2.8394
Epoch 46/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:29 7s/step - accuracy: 0.5312 - loss: 2.8342
Epoch 46: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 91ms/step - accuracy: 0.5312 - loss: 2.8342 - val_accuracy: 0.4412 - val_loss: 2.8315
Epoch 47/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5163 - loss: 2.8347
Epoch 47: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.5162 - loss: 2.8347 - val_accuracy: 0.4943 - val_loss: 2.8352
Epoch 48/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.5000 - loss: 2.8295
Epoch 48: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 91ms/step - accuracy: 0.5000 - loss: 2.8295 - val_accuracy: 0.5294 - val_loss: 2.8346
Epoch 49/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4969 - loss: 2.8336
Epoch 49: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.4971 - loss: 2.8335 - val_accuracy: 0.4972 - val_loss: 2.8331
Epoch 50/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.5625 - loss: 2.8222
Epoch 50: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.5625 - loss: 2.8222 - val_accuracy: 0.4118 - val_loss: 2.8404
Restoring model weights from the end of the best epoch: 42.
Tiempo transcurrido para el entrenamiento: 8758.206644535065 segundos
Uso de CPU durante el entrenamiento: 44.7%
Aumento en uso de memoria: 0.4933815002441406 GB
Resultados para lr=0.0001, l2=0.1, batch_size=64:
Tiempo: 8758.206644535065 segundos, CPU: 44.7%, Memoria: 0.4933815002441406 GB
Precisión de validación: 0.6176470518112183
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_22"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_11 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.2015 - loss: 3.0636
Epoch 1: val_accuracy improved from -inf to 0.42663, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 360s 2s/step - accuracy: 0.2020 - loss: 3.0630 - val_accuracy: 0.4266 - val_loss: 2.8548
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 5:51 2s/step - accuracy: 0.4375 - loss: 2.8509
Epoch 2: val_accuracy improved from 0.42663 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 5s 18ms/step - accuracy: 0.4375 - loss: 2.8509 - val_accuracy: 1.0000 - val_loss: 2.8538
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4730 - loss: 2.8419
Epoch 3: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 358s 2s/step - accuracy: 0.4730 - loss: 2.8419 - val_accuracy: 0.5245 - val_loss: 2.8181
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.5625 - loss: 2.8304
Epoch 4: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5625 - loss: 2.8304 - val_accuracy: 0.5000 - val_loss: 2.7663
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5449 - loss: 2.8118
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.5449 - loss: 2.8118 - val_accuracy: 0.5299 - val_loss: 2.7961
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.5625 - loss: 2.7999
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5625 - loss: 2.7999 - val_accuracy: 0.5000 - val_loss: 2.8246
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5746 - loss: 2.7888
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.5746 - loss: 2.7888 - val_accuracy: 0.5408 - val_loss: 2.7788
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.5625 - loss: 2.7803
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5625 - loss: 2.7803 - val_accuracy: 0.5000 - val_loss: 2.8608
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5752 - loss: 2.7687
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.5752 - loss: 2.7687 - val_accuracy: 0.5747 - val_loss: 2.7654
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.7500 - loss: 2.7698
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.7500 - loss: 2.7698 - val_accuracy: 1.0000 - val_loss: 2.7166
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5964 - loss: 2.7587
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.5964 - loss: 2.7586 - val_accuracy: 0.5611 - val_loss: 2.7552
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.5625 - loss: 2.7537
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5625 - loss: 2.7537 - val_accuracy: 0.0000e+00 - val_loss: 2.8822
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2132.980417728424 segundos
Uso de CPU durante el entrenamiento: 37.8%
Aumento en uso de memoria: -1.5767440795898438 GB
Resultados para lr=0.0005, l2=0.01, batch_size=16:
Tiempo: 2132.980417728424 segundos, CPU: 37.8%, Memoria: -1.5767440795898438 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_24"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_12 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.1225 - loss: 3.1285
Epoch 1: val_accuracy improved from -inf to 0.35734, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 349s 4s/step - accuracy: 0.1237 - loss: 3.1275 - val_accuracy: 0.3573 - val_loss: 2.9184
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:09 3s/step - accuracy: 0.3750 - loss: 2.9194
Epoch 2: val_accuracy did not improve from 0.35734
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.3750 - loss: 2.9194 - val_accuracy: 0.0000e+00 - val_loss: 2.9299
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.4026 - loss: 2.8914
Epoch 3: val_accuracy improved from 0.35734 to 0.47283, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 346s 4s/step - accuracy: 0.4028 - loss: 2.8912 - val_accuracy: 0.4728 - val_loss: 2.8442
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 5:21 4s/step - accuracy: 0.5000 - loss: 2.8305
Epoch 4: val_accuracy improved from 0.47283 to 0.50000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 6s 28ms/step - accuracy: 0.5000 - loss: 2.8305 - val_accuracy: 0.5000 - val_loss: 2.8547
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.4857 - loss: 2.8375
Epoch 5: val_accuracy improved from 0.50000 to 0.52717, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 350s 4s/step - accuracy: 0.4858 - loss: 2.8374 - val_accuracy: 0.5272 - val_loss: 2.8199
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:59 3s/step - accuracy: 0.5938 - loss: 2.7965
Epoch 6: val_accuracy did not improve from 0.52717
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5938 - loss: 2.7965 - val_accuracy: 0.5000 - val_loss: 2.8333
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5424 - loss: 2.8162
Epoch 7: val_accuracy improved from 0.52717 to 0.54348, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 349s 4s/step - accuracy: 0.5422 - loss: 2.8162 - val_accuracy: 0.5435 - val_loss: 2.8047
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:01 3s/step - accuracy: 0.6562 - loss: 2.8174
Epoch 8: val_accuracy did not improve from 0.54348
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6562 - loss: 2.8174 - val_accuracy: 0.0000e+00 - val_loss: 2.8058
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5433 - loss: 2.8015
Epoch 9: val_accuracy improved from 0.54348 to 0.57337, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.5434 - loss: 2.8015 - val_accuracy: 0.5734 - val_loss: 2.7906
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.5625 - loss: 2.7664
Epoch 10: val_accuracy did not improve from 0.57337
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5625 - loss: 2.7664 - val_accuracy: 0.5000 - val_loss: 2.7853
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5571 - loss: 2.7870
Epoch 11: val_accuracy improved from 0.57337 to 0.57609, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.5572 - loss: 2.7870 - val_accuracy: 0.5761 - val_loss: 2.7829
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.5312 - loss: 2.7984
Epoch 12: val_accuracy improved from 0.57609 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 5s 30ms/step - accuracy: 0.5312 - loss: 2.7984 - val_accuracy: 1.0000 - val_loss: 2.6973
Epoch 13/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5623 - loss: 2.7765
Epoch 13: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.5623 - loss: 2.7765 - val_accuracy: 0.5693 - val_loss: 2.7741
Epoch 14/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:04 3s/step - accuracy: 0.5312 - loss: 2.7777
Epoch 14: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 5ms/step - accuracy: 0.5312 - loss: 2.7777 - val_accuracy: 0.5000 - val_loss: 2.8106
Epoch 15/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5682 - loss: 2.7690
Epoch 15: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.5682 - loss: 2.7689 - val_accuracy: 0.5611 - val_loss: 2.7630
Epoch 16/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 5:14 4s/step - accuracy: 0.5625 - loss: 2.7736
Epoch 16: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.5625 - loss: 2.7736 - val_accuracy: 1.0000 - val_loss: 2.7360
Epoch 17/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5740 - loss: 2.7564
Epoch 17: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 349s 4s/step - accuracy: 0.5739 - loss: 2.7564 - val_accuracy: 0.5557 - val_loss: 2.7505
Epoch 18/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.5625 - loss: 2.7368
Epoch 18: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5625 - loss: 2.7368 - val_accuracy: 0.0000e+00 - val_loss: 2.6997
Epoch 19/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5686 - loss: 2.7498
Epoch 19: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.5686 - loss: 2.7498 - val_accuracy: 0.5666 - val_loss: 2.7487
Epoch 20/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:04 3s/step - accuracy: 0.6562 - loss: 2.7347
Epoch 20: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6562 - loss: 2.7347 - val_accuracy: 0.5000 - val_loss: 2.7975
Epoch 21/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5831 - loss: 2.7454
Epoch 21: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 341s 4s/step - accuracy: 0.5830 - loss: 2.7454 - val_accuracy: 0.5421 - val_loss: 2.7407
Epoch 22/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:01 3s/step - accuracy: 0.5625 - loss: 2.7088
Epoch 22: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5625 - loss: 2.7088 - val_accuracy: 0.5000 - val_loss: 2.8002
Epoch 22: early stopping
Restoring model weights from the end of the best epoch: 12.
Tiempo transcurrido para el entrenamiento: 3846.1872708797455 segundos
Uso de CPU durante el entrenamiento: 44.00000000000001%
Aumento en uso de memoria: -0.29264068603515625 GB
Resultados para lr=0.0005, l2=0.01, batch_size=32:
Tiempo: 3846.1872708797455 segundos, CPU: 44.00000000000001%, Memoria: -0.29264068603515625 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_26"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_13 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.0913 - loss: 3.1883
Epoch 1: val_accuracy improved from -inf to 0.19034, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 353s 9s/step - accuracy: 0.0921 - loss: 3.1870 - val_accuracy: 0.1903 - val_loss: 3.0344
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:56 8s/step - accuracy: 0.1875 - loss: 3.0358
Epoch 2: val_accuracy improved from 0.19034 to 0.23529, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 14s 165ms/step - accuracy: 0.1875 - loss: 3.0358 - val_accuracy: 0.2353 - val_loss: 3.0330
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.2273 - loss: 2.9999
Epoch 3: val_accuracy improved from 0.23529 to 0.29261, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 341s 8s/step - accuracy: 0.2279 - loss: 2.9993 - val_accuracy: 0.2926 - val_loss: 2.9268
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.4062 - loss: 2.9223
Epoch 4: val_accuracy did not improve from 0.29261
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 92ms/step - accuracy: 0.4062 - loss: 2.9223 - val_accuracy: 0.2647 - val_loss: 2.9212
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.3170 - loss: 2.9104
Epoch 5: val_accuracy improved from 0.29261 to 0.39205, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 344s 9s/step - accuracy: 0.3177 - loss: 2.9101 - val_accuracy: 0.3920 - val_loss: 2.8765
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.2812 - loss: 2.8816
Epoch 6: val_accuracy improved from 0.39205 to 0.41176, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 171ms/step - accuracy: 0.2812 - loss: 2.8816 - val_accuracy: 0.4118 - val_loss: 2.8821
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4059 - loss: 2.8680
Epoch 7: val_accuracy improved from 0.41176 to 0.42472, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 346s 9s/step - accuracy: 0.4060 - loss: 2.8679 - val_accuracy: 0.4247 - val_loss: 2.8526
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.5312 - loss: 2.8458
Epoch 8: val_accuracy improved from 0.42472 to 0.47059, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 136ms/step - accuracy: 0.5312 - loss: 2.8458 - val_accuracy: 0.4706 - val_loss: 2.8424
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4369 - loss: 2.8461
Epoch 9: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.4372 - loss: 2.8461 - val_accuracy: 0.4432 - val_loss: 2.8361
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:10 6s/step - accuracy: 0.4375 - loss: 2.8359
Epoch 10: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.4375 - loss: 2.8359 - val_accuracy: 0.4706 - val_loss: 2.8332
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4317 - loss: 2.8323
Epoch 11: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.4324 - loss: 2.8323 - val_accuracy: 0.4517 - val_loss: 2.8266
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:17 7s/step - accuracy: 0.4531 - loss: 2.8205
Epoch 12: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.4531 - loss: 2.8205 - val_accuracy: 0.4118 - val_loss: 2.8277
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4823 - loss: 2.8222
Epoch 13: val_accuracy improved from 0.47059 to 0.48438, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.4822 - loss: 2.8222 - val_accuracy: 0.4844 - val_loss: 2.8176
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:55 6s/step - accuracy: 0.5156 - loss: 2.8041
Epoch 14: val_accuracy improved from 0.48438 to 0.58824, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 138ms/step - accuracy: 0.5156 - loss: 2.8041 - val_accuracy: 0.5882 - val_loss: 2.8088
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5059 - loss: 2.8132
Epoch 15: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.5057 - loss: 2.8131 - val_accuracy: 0.4886 - val_loss: 2.8099
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:12 6s/step - accuracy: 0.5625 - loss: 2.8102
Epoch 16: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.5625 - loss: 2.8102 - val_accuracy: 0.3235 - val_loss: 2.8336
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4785 - loss: 2.8060
Epoch 17: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.4788 - loss: 2.8059 - val_accuracy: 0.4815 - val_loss: 2.8019
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.5469 - loss: 2.8090
Epoch 18: val_accuracy improved from 0.58824 to 0.61765, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 154ms/step - accuracy: 0.5469 - loss: 2.8090 - val_accuracy: 0.6176 - val_loss: 2.8123
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5063 - loss: 2.7978
Epoch 19: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.5062 - loss: 2.7978 - val_accuracy: 0.4957 - val_loss: 2.7941
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.5469 - loss: 2.7938
Epoch 20: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.5469 - loss: 2.7938 - val_accuracy: 0.4412 - val_loss: 2.7993
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4973 - loss: 2.7926
Epoch 21: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 345s 9s/step - accuracy: 0.4973 - loss: 2.7925 - val_accuracy: 0.5014 - val_loss: 2.7900
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.5469 - loss: 2.7896
Epoch 22: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 99ms/step - accuracy: 0.5469 - loss: 2.7896 - val_accuracy: 0.4412 - val_loss: 2.8018
Epoch 23/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5162 - loss: 2.7847
Epoch 23: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.5162 - loss: 2.7846 - val_accuracy: 0.5156 - val_loss: 2.7841
Epoch 24/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.6094 - loss: 2.7578
Epoch 24: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.6094 - loss: 2.7578 - val_accuracy: 0.5882 - val_loss: 2.8058
Epoch 25/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5063 - loss: 2.7786
Epoch 25: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5064 - loss: 2.7785 - val_accuracy: 0.5156 - val_loss: 2.7765
Epoch 26/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:53 8s/step - accuracy: 0.5469 - loss: 2.7863
Epoch 26: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 100ms/step - accuracy: 0.5469 - loss: 2.7863 - val_accuracy: 0.4412 - val_loss: 2.7992
Epoch 27/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5200 - loss: 2.7737
Epoch 27: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.5200 - loss: 2.7736 - val_accuracy: 0.4929 - val_loss: 2.7722
Epoch 28/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:05 6s/step - accuracy: 0.4844 - loss: 2.7669
Epoch 28: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.4844 - loss: 2.7669 - val_accuracy: 0.5294 - val_loss: 2.7917
Epoch 28: early stopping
Restoring model weights from the end of the best epoch: 18.
Tiempo transcurrido para el entrenamiento: 4920.667220592499 segundos
Uso de CPU durante el entrenamiento: 32.10000000000001%
Aumento en uso de memoria: 0.7705421447753906 GB
Resultados para lr=0.0005, l2=0.01, batch_size=64:
Tiempo: 4920.667220592499 segundos, CPU: 32.10000000000001%, Memoria: 0.7705421447753906 GB
Precisión de validación: 0.6176470518112183
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_28"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_14 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.1560 - loss: 3.0733
Epoch 1: val_accuracy improved from -inf to 0.40217, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 360s 2s/step - accuracy: 0.1565 - loss: 3.0727 - val_accuracy: 0.4022 - val_loss: 2.8613
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:58 2s/step - accuracy: 0.4375 - loss: 2.8690
Epoch 2: val_accuracy improved from 0.40217 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 5s 20ms/step - accuracy: 0.4375 - loss: 2.8690 - val_accuracy: 1.0000 - val_loss: 2.8305
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4571 - loss: 2.8489
Epoch 3: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 355s 2s/step - accuracy: 0.4573 - loss: 2.8489 - val_accuracy: 0.5326 - val_loss: 2.8251
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:33 2s/step - accuracy: 0.6250 - loss: 2.8501
Epoch 4: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6250 - loss: 2.8501 - val_accuracy: 0.5000 - val_loss: 2.8070
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5417 - loss: 2.8167
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 365s 2s/step - accuracy: 0.5419 - loss: 2.8166 - val_accuracy: 0.5584 - val_loss: 2.8020
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:20 2s/step - accuracy: 0.5000 - loss: 2.7923
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5000 - loss: 2.7923 - val_accuracy: 1.0000 - val_loss: 2.8346
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5843 - loss: 2.7952
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.5843 - loss: 2.7952 - val_accuracy: 0.5611 - val_loss: 2.7875
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.7500 - loss: 2.7362
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.7500 - loss: 2.7362 - val_accuracy: 1.0000 - val_loss: 2.7071
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5714 - loss: 2.7773
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.5715 - loss: 2.7773 - val_accuracy: 0.5910 - val_loss: 2.7708
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.5000 - loss: 2.7910
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5000 - loss: 2.7910 - val_accuracy: 0.5000 - val_loss: 2.8868
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5866 - loss: 2.7607
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 352s 2s/step - accuracy: 0.5865 - loss: 2.7607 - val_accuracy: 0.5611 - val_loss: 2.7561
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 5:28 2s/step - accuracy: 0.5625 - loss: 2.8015
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5625 - loss: 2.8015 - val_accuracy: 0.0000e+00 - val_loss: 2.7754
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2145.9741504192352 segundos
Uso de CPU durante el entrenamiento: 58.9%
Aumento en uso de memoria: -1.9600105285644531 GB
Resultados para lr=0.0005, l2=0.05, batch_size=16:
Tiempo: 2145.9741504192352 segundos, CPU: 58.9%, Memoria: -1.9600105285644531 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_30"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_15 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.1232 - loss: 3.1322
Epoch 1: val_accuracy improved from -inf to 0.30707, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 353s 4s/step - accuracy: 0.1240 - loss: 3.1312 - val_accuracy: 0.3071 - val_loss: 2.9228
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.4062 - loss: 2.9174
Epoch 2: val_accuracy did not improve from 0.30707
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.4062 - loss: 2.9174 - val_accuracy: 0.0000e+00 - val_loss: 2.9656
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.3718 - loss: 2.8941
Epoch 3: val_accuracy improved from 0.30707 to 0.41304, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 347s 4s/step - accuracy: 0.3720 - loss: 2.8939 - val_accuracy: 0.4130 - val_loss: 2.8485
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:13 3s/step - accuracy: 0.4375 - loss: 2.8420
Epoch 4: val_accuracy did not improve from 0.41304
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.4375 - loss: 2.8420 - val_accuracy: 0.0000e+00 - val_loss: 2.8923
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.4353 - loss: 2.8410
Epoch 5: val_accuracy improved from 0.41304 to 0.48505, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.4356 - loss: 2.8410 - val_accuracy: 0.4851 - val_loss: 2.8258
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.4062 - loss: 2.8252
Epoch 6: val_accuracy did not improve from 0.48505
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.4062 - loss: 2.8252 - val_accuracy: 0.0000e+00 - val_loss: 2.8099
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5247 - loss: 2.8197
Epoch 7: val_accuracy improved from 0.48505 to 0.50272, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 357s 4s/step - accuracy: 0.5248 - loss: 2.8196 - val_accuracy: 0.5027 - val_loss: 2.8085
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.7188 - loss: 2.8131
Epoch 8: val_accuracy did not improve from 0.50272
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.7188 - loss: 2.8131 - val_accuracy: 0.5000 - val_loss: 2.8012
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5326 - loss: 2.8040
Epoch 9: val_accuracy improved from 0.50272 to 0.51766, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.5329 - loss: 2.8039 - val_accuracy: 0.5177 - val_loss: 2.7982
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.6562 - loss: 2.7964
Epoch 10: val_accuracy improved from 0.51766 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 5s 27ms/step - accuracy: 0.6562 - loss: 2.7964 - val_accuracy: 1.0000 - val_loss: 2.7540
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5706 - loss: 2.7896
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.5705 - loss: 2.7896 - val_accuracy: 0.5448 - val_loss: 2.7854
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:12 3s/step - accuracy: 0.6875 - loss: 2.7879
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6875 - loss: 2.7879 - val_accuracy: 0.0000e+00 - val_loss: 2.7511
Epoch 13/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5848 - loss: 2.7782
Epoch 13: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 341s 4s/step - accuracy: 0.5847 - loss: 2.7782 - val_accuracy: 0.5639 - val_loss: 2.7797
Epoch 14/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.6562 - loss: 2.7627
Epoch 14: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6562 - loss: 2.7627 - val_accuracy: 0.5000 - val_loss: 2.7438
Epoch 15/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.6072 - loss: 2.7692
Epoch 15: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.6070 - loss: 2.7692 - val_accuracy: 0.5774 - val_loss: 2.7682
Epoch 16/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:29 3s/step - accuracy: 0.6250 - loss: 2.7734
Epoch 16: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.6250 - loss: 2.7734 - val_accuracy: 0.0000e+00 - val_loss: 2.7860
Epoch 17/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5975 - loss: 2.7622
Epoch 17: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.5975 - loss: 2.7622 - val_accuracy: 0.5815 - val_loss: 2.7582
Epoch 18/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.5938 - loss: 2.7760
Epoch 18: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5938 - loss: 2.7760 - val_accuracy: 1.0000 - val_loss: 2.6446
Epoch 19/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.6032 - loss: 2.7521
Epoch 19: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.6031 - loss: 2.7521 - val_accuracy: 0.5870 - val_loss: 2.7540
Epoch 20/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.5938 - loss: 2.7687
Epoch 20: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5938 - loss: 2.7687 - val_accuracy: 0.5000 - val_loss: 2.7359
Epoch 20: early stopping
Restoring model weights from the end of the best epoch: 10.
Tiempo transcurrido para el entrenamiento: 3494.3435051441193 segundos
Uso de CPU durante el entrenamiento: 43.3%
Aumento en uso de memoria: 0.5475082397460938 GB
Resultados para lr=0.0005, l2=0.05, batch_size=32:
Tiempo: 3494.3435051441193 segundos, CPU: 43.3%, Memoria: 0.5475082397460938 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_32"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_16 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.0984 - loss: 3.1908
Epoch 1: val_accuracy improved from -inf to 0.18466, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 353s 9s/step - accuracy: 0.0990 - loss: 3.1895 - val_accuracy: 0.1847 - val_loss: 3.0370
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.1562 - loss: 3.0350
Epoch 2: val_accuracy improved from 0.18466 to 0.26471, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 173ms/step - accuracy: 0.1562 - loss: 3.0350 - val_accuracy: 0.2647 - val_loss: 3.0281
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.1981 - loss: 3.0008
Epoch 3: val_accuracy improved from 0.26471 to 0.28977, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 345s 9s/step - accuracy: 0.1987 - loss: 3.0002 - val_accuracy: 0.2898 - val_loss: 2.9268
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:23 7s/step - accuracy: 0.2188 - loss: 2.9286
Epoch 4: val_accuracy did not improve from 0.28977
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 101ms/step - accuracy: 0.2188 - loss: 2.9286 - val_accuracy: 0.2647 - val_loss: 2.9245
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.3239 - loss: 2.9095
Epoch 5: val_accuracy improved from 0.28977 to 0.35795, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.3241 - loss: 2.9092 - val_accuracy: 0.3580 - val_loss: 2.8748
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.3438 - loss: 2.8800
Epoch 6: val_accuracy improved from 0.35795 to 0.41176, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 147ms/step - accuracy: 0.3438 - loss: 2.8800 - val_accuracy: 0.4118 - val_loss: 2.8793
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.3991 - loss: 2.8673
Epoch 7: val_accuracy improved from 0.41176 to 0.46023, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 343s 8s/step - accuracy: 0.3997 - loss: 2.8672 - val_accuracy: 0.4602 - val_loss: 2.8506
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.4688 - loss: 2.8462
Epoch 8: val_accuracy did not improve from 0.46023
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 90ms/step - accuracy: 0.4688 - loss: 2.8462 - val_accuracy: 0.2941 - val_loss: 2.8561
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4757 - loss: 2.8448
Epoch 9: val_accuracy improved from 0.46023 to 0.50710, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.4760 - loss: 2.8447 - val_accuracy: 0.5071 - val_loss: 2.8354
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.5312 - loss: 2.8394
Epoch 10: val_accuracy improved from 0.50710 to 0.58824, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 140ms/step - accuracy: 0.5312 - loss: 2.8394 - val_accuracy: 0.5882 - val_loss: 2.8389
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5162 - loss: 2.8327
Epoch 11: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5164 - loss: 2.8326 - val_accuracy: 0.5185 - val_loss: 2.8238
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:13 6s/step - accuracy: 0.5312 - loss: 2.8164
Epoch 12: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 90ms/step - accuracy: 0.5312 - loss: 2.8164 - val_accuracy: 0.4412 - val_loss: 2.8324
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5300 - loss: 2.8204
Epoch 13: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 343s 8s/step - accuracy: 0.5298 - loss: 2.8203 - val_accuracy: 0.5241 - val_loss: 2.8141
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:14 7s/step - accuracy: 0.5000 - loss: 2.8074
Epoch 14: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 98ms/step - accuracy: 0.5000 - loss: 2.8074 - val_accuracy: 0.4118 - val_loss: 2.8212
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5559 - loss: 2.8102
Epoch 15: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5555 - loss: 2.8102 - val_accuracy: 0.5526 - val_loss: 2.8066
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.5000 - loss: 2.8009
Epoch 16: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 95ms/step - accuracy: 0.5000 - loss: 2.8009 - val_accuracy: 0.5882 - val_loss: 2.8188
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5608 - loss: 2.8000
Epoch 17: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5605 - loss: 2.8000 - val_accuracy: 0.5426 - val_loss: 2.7993
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.5625 - loss: 2.7925
Epoch 18: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.5625 - loss: 2.7925 - val_accuracy: 0.5000 - val_loss: 2.8058
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5456 - loss: 2.7949
Epoch 19: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.5460 - loss: 2.7949 - val_accuracy: 0.5582 - val_loss: 2.7939
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.5156 - loss: 2.7990
Epoch 20: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.5156 - loss: 2.7990 - val_accuracy: 0.4706 - val_loss: 2.7889
Epoch 20: early stopping
Restoring model weights from the end of the best epoch: 10.
Tiempo transcurrido para el entrenamiento: 3517.071481704712 segundos
Uso de CPU durante el entrenamiento: 45.900000000000006%
Aumento en uso de memoria: 0.386474609375 GB
Resultados para lr=0.0005, l2=0.05, batch_size=64:
Tiempo: 3517.071481704712 segundos, CPU: 45.900000000000006%, Memoria: 0.386474609375 GB
Precisión de validación: 0.5882353186607361
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_34"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_17 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.1224 - loss: 3.0702
Epoch 1: val_accuracy improved from -inf to 0.37636, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 357s 2s/step - accuracy: 0.1229 - loss: 3.0695 - val_accuracy: 0.3764 - val_loss: 2.8578
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:28 2s/step - accuracy: 0.3750 - loss: 2.8611
Epoch 2: val_accuracy improved from 0.37636 to 0.50000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 4s 16ms/step - accuracy: 0.3750 - loss: 2.8611 - val_accuracy: 0.5000 - val_loss: 2.8441
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4033 - loss: 2.8455
Epoch 3: val_accuracy improved from 0.50000 to 0.51223, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 354s 2s/step - accuracy: 0.4035 - loss: 2.8454 - val_accuracy: 0.5122 - val_loss: 2.8205
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.3750 - loss: 2.8013
Epoch 4: val_accuracy did not improve from 0.51223
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3750 - loss: 2.8013 - val_accuracy: 0.5000 - val_loss: 2.8679
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5420 - loss: 2.8140
Epoch 5: val_accuracy improved from 0.51223 to 0.55707, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 357s 2s/step - accuracy: 0.5420 - loss: 2.8140 - val_accuracy: 0.5571 - val_loss: 2.8003
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 5:18 2s/step - accuracy: 0.6875 - loss: 2.7869
Epoch 6: val_accuracy did not improve from 0.55707
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6875 - loss: 2.7869 - val_accuracy: 0.5000 - val_loss: 2.7053
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5787 - loss: 2.7938
Epoch 7: val_accuracy improved from 0.55707 to 0.57337, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 360s 2s/step - accuracy: 0.5787 - loss: 2.7937 - val_accuracy: 0.5734 - val_loss: 2.7830
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 0.3750 - loss: 2.7758
Epoch 8: val_accuracy did not improve from 0.57337
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3750 - loss: 2.7758 - val_accuracy: 0.5000 - val_loss: 2.7883
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5754 - loss: 2.7747
Epoch 9: val_accuracy did not improve from 0.57337
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.5753 - loss: 2.7747 - val_accuracy: 0.5584 - val_loss: 2.7701
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:15 2s/step - accuracy: 0.4375 - loss: 2.7617
Epoch 10: val_accuracy improved from 0.57337 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 4s 14ms/step - accuracy: 0.4375 - loss: 2.7617 - val_accuracy: 1.0000 - val_loss: 2.7330
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5818 - loss: 2.7604
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 353s 2s/step - accuracy: 0.5819 - loss: 2.7604 - val_accuracy: 0.5720 - val_loss: 2.7556
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.7500 - loss: 2.7402
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.7500 - loss: 2.7402 - val_accuracy: 1.0000 - val_loss: 2.8124
Epoch 13/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5780 - loss: 2.7522
Epoch 13: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.5780 - loss: 2.7522 - val_accuracy: 0.5842 - val_loss: 2.7451
Epoch 14/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.6250 - loss: 2.7075
Epoch 14: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6250 - loss: 2.7075 - val_accuracy: 0.5000 - val_loss: 2.8552
Epoch 15/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5934 - loss: 2.7420
Epoch 15: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.5935 - loss: 2.7419 - val_accuracy: 0.5734 - val_loss: 2.7335
Epoch 16/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.6250 - loss: 2.7029
Epoch 16: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6250 - loss: 2.7029 - val_accuracy: 0.5000 - val_loss: 2.7628
Epoch 17/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5761 - loss: 2.7330
Epoch 17: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 359s 2s/step - accuracy: 0.5761 - loss: 2.7330 - val_accuracy: 0.5856 - val_loss: 2.7300
Epoch 18/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.6875 - loss: 2.7521
Epoch 18: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6875 - loss: 2.7521 - val_accuracy: 1.0000 - val_loss: 2.6692
Epoch 19/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5901 - loss: 2.7168
Epoch 19: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.5900 - loss: 2.7168 - val_accuracy: 0.5897 - val_loss: 2.7216
Epoch 20/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.3750 - loss: 2.7470
Epoch 20: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.3750 - loss: 2.7470 - val_accuracy: 0.5000 - val_loss: 2.9133
Epoch 20: early stopping
Restoring model weights from the end of the best epoch: 10.
Tiempo transcurrido para el entrenamiento: 3565.7795617580414 segundos
Uso de CPU durante el entrenamiento: 45.599999999999994%
Aumento en uso de memoria: -0.4730491638183594 GB
Resultados para lr=0.0005, l2=0.1, batch_size=16:
Tiempo: 3565.7795617580414 segundos, CPU: 45.599999999999994%, Memoria: -0.4730491638183594 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_36"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_18 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.1433 - loss: 3.1293
Epoch 1: val_accuracy improved from -inf to 0.33967, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 352s 4s/step - accuracy: 0.1442 - loss: 3.1283 - val_accuracy: 0.3397 - val_loss: 2.9191
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.3750 - loss: 2.9157
Epoch 2: val_accuracy improved from 0.33967 to 0.50000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 6s 35ms/step - accuracy: 0.3750 - loss: 2.9157 - val_accuracy: 0.5000 - val_loss: 2.9243
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.4088 - loss: 2.8918
Epoch 3: val_accuracy did not improve from 0.50000
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.4090 - loss: 2.8915 - val_accuracy: 0.4633 - val_loss: 2.8461
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:12 3s/step - accuracy: 0.3793 - loss: 2.8483
Epoch 4: val_accuracy improved from 0.50000 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 5s 29ms/step - accuracy: 0.3793 - loss: 2.8483 - val_accuracy: 1.0000 - val_loss: 2.8113
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.4844 - loss: 2.8378
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.4845 - loss: 2.8377 - val_accuracy: 0.4946 - val_loss: 2.8223
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.4688 - loss: 2.8319
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.4688 - loss: 2.8319 - val_accuracy: 0.5000 - val_loss: 2.7750
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5468 - loss: 2.8173
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 349s 4s/step - accuracy: 0.5466 - loss: 2.8173 - val_accuracy: 0.5299 - val_loss: 2.8081
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:19 3s/step - accuracy: 0.5000 - loss: 2.8177
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 6ms/step - accuracy: 0.5000 - loss: 2.8177 - val_accuracy: 0.5000 - val_loss: 2.7218
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5403 - loss: 2.8016
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.5405 - loss: 2.8015 - val_accuracy: 0.5353 - val_loss: 2.7943
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.4688 - loss: 2.7928
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.4688 - loss: 2.7928 - val_accuracy: 1.0000 - val_loss: 2.8047
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5486 - loss: 2.7937
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 349s 4s/step - accuracy: 0.5486 - loss: 2.7936 - val_accuracy: 0.5516 - val_loss: 2.7839
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:17 3s/step - accuracy: 0.6562 - loss: 2.7864
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.6562 - loss: 2.7864 - val_accuracy: 1.0000 - val_loss: 2.7364
Epoch 13/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5737 - loss: 2.7767
Epoch 13: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.5737 - loss: 2.7767 - val_accuracy: 0.5530 - val_loss: 2.7738
Epoch 14/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:26 3s/step - accuracy: 0.6562 - loss: 2.7798
Epoch 14: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.6562 - loss: 2.7798 - val_accuracy: 0.5000 - val_loss: 2.8315
Epoch 14: early stopping
Restoring model weights from the end of the best epoch: 4.
Tiempo transcurrido para el entrenamiento: 2456.2960608005524 segundos
Uso de CPU durante el entrenamiento: 49.599999999999994%
Aumento en uso de memoria: -0.13779067993164062 GB
Resultados para lr=0.0005, l2=0.1, batch_size=32:
Tiempo: 2456.2960608005524 segundos, CPU: 49.599999999999994%, Memoria: -0.13779067993164062 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_38"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_19 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.0843 - loss: 3.1845
Epoch 1: val_accuracy improved from -inf to 0.16335, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 345s 8s/step - accuracy: 0.0848 - loss: 3.1832 - val_accuracy: 0.1634 - val_loss: 3.0301
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.1719 - loss: 3.0297
Epoch 2: val_accuracy did not improve from 0.16335
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 90ms/step - accuracy: 0.1719 - loss: 3.0297 - val_accuracy: 0.1471 - val_loss: 3.0253
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.2015 - loss: 2.9939
Epoch 3: val_accuracy improved from 0.16335 to 0.32386, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 343s 8s/step - accuracy: 0.2021 - loss: 2.9932 - val_accuracy: 0.3239 - val_loss: 2.9209
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:07 6s/step - accuracy: 0.2031 - loss: 2.9261
Epoch 4: val_accuracy did not improve from 0.32386
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 99ms/step - accuracy: 0.2031 - loss: 2.9261 - val_accuracy: 0.2353 - val_loss: 2.9240
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.3370 - loss: 2.9046
Epoch 5: val_accuracy improved from 0.32386 to 0.39915, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.3376 - loss: 2.9042 - val_accuracy: 0.3991 - val_loss: 2.8710
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:07 6s/step - accuracy: 0.3438 - loss: 2.8682
Epoch 6: val_accuracy did not improve from 0.39915
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 95ms/step - accuracy: 0.3438 - loss: 2.8682 - val_accuracy: 0.2353 - val_loss: 2.8756
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4249 - loss: 2.8615
Epoch 7: val_accuracy improved from 0.39915 to 0.43040, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 347s 9s/step - accuracy: 0.4249 - loss: 2.8613 - val_accuracy: 0.4304 - val_loss: 2.8443
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:25 7s/step - accuracy: 0.5156 - loss: 2.8395
Epoch 8: val_accuracy improved from 0.43040 to 0.47059, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 14s 189ms/step - accuracy: 0.5156 - loss: 2.8395 - val_accuracy: 0.4706 - val_loss: 2.8478
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4604 - loss: 2.8401
Epoch 9: val_accuracy did not improve from 0.47059
40/40 ━━━━━━━━━━━━━━━━━━━━ 361s 9s/step - accuracy: 0.4605 - loss: 2.8400 - val_accuracy: 0.4645 - val_loss: 2.8300
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:33 7s/step - accuracy: 0.5781 - loss: 2.8216
Epoch 10: val_accuracy improved from 0.47059 to 0.50000, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 165ms/step - accuracy: 0.5781 - loss: 2.8216 - val_accuracy: 0.5000 - val_loss: 2.8279
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4707 - loss: 2.8250
Epoch 11: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 360s 9s/step - accuracy: 0.4712 - loss: 2.8250 - val_accuracy: 0.4787 - val_loss: 2.8187
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.4844 - loss: 2.8225
Epoch 12: val_accuracy improved from 0.50000 to 0.52941, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 163ms/step - accuracy: 0.4844 - loss: 2.8225 - val_accuracy: 0.5294 - val_loss: 2.8061
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5207 - loss: 2.8128
Epoch 13: val_accuracy did not improve from 0.52941
40/40 ━━━━━━━━━━━━━━━━━━━━ 345s 9s/step - accuracy: 0.5201 - loss: 2.8127 - val_accuracy: 0.4858 - val_loss: 2.8107
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:32 7s/step - accuracy: 0.4688 - loss: 2.8220
Epoch 14: val_accuracy improved from 0.52941 to 0.55882, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 160ms/step - accuracy: 0.4688 - loss: 2.8220 - val_accuracy: 0.5588 - val_loss: 2.8144
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5127 - loss: 2.8049
Epoch 15: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.5126 - loss: 2.8048 - val_accuracy: 0.5142 - val_loss: 2.8010
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.5156 - loss: 2.7943
Epoch 16: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.5156 - loss: 2.7943 - val_accuracy: 0.4706 - val_loss: 2.8339
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5338 - loss: 2.7985
Epoch 17: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.5334 - loss: 2.7985 - val_accuracy: 0.5114 - val_loss: 2.7965
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.5000 - loss: 2.7846
Epoch 18: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.5000 - loss: 2.7846 - val_accuracy: 0.4118 - val_loss: 2.7912
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5354 - loss: 2.7922
Epoch 19: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.5352 - loss: 2.7921 - val_accuracy: 0.5128 - val_loss: 2.7891
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:10 6s/step - accuracy: 0.5312 - loss: 2.7828
Epoch 20: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.5312 - loss: 2.7828 - val_accuracy: 0.5294 - val_loss: 2.7834
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5534 - loss: 2.7845
Epoch 21: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.5528 - loss: 2.7845 - val_accuracy: 0.5497 - val_loss: 2.7817
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.5625 - loss: 2.7924
Epoch 22: val_accuracy improved from 0.55882 to 0.58824, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 179ms/step - accuracy: 0.5625 - loss: 2.7924 - val_accuracy: 0.5882 - val_loss: 2.7930
Epoch 23/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5506 - loss: 2.7798
Epoch 23: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 346s 9s/step - accuracy: 0.5504 - loss: 2.7798 - val_accuracy: 0.5355 - val_loss: 2.7753
Epoch 24/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:12 6s/step - accuracy: 0.5938 - loss: 2.7751
Epoch 24: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.5938 - loss: 2.7751 - val_accuracy: 0.5882 - val_loss: 2.7581
Epoch 25/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5620 - loss: 2.7731
Epoch 25: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.5617 - loss: 2.7731 - val_accuracy: 0.5369 - val_loss: 2.7733
Epoch 26/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:12 6s/step - accuracy: 0.6094 - loss: 2.7689
Epoch 26: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 92ms/step - accuracy: 0.6094 - loss: 2.7689 - val_accuracy: 0.5588 - val_loss: 2.7548
Epoch 27/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5503 - loss: 2.7677
Epoch 27: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.5507 - loss: 2.7676 - val_accuracy: 0.5753 - val_loss: 2.7673
Epoch 28/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.6250 - loss: 2.7754
Epoch 28: val_accuracy improved from 0.58824 to 0.64706, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 171ms/step - accuracy: 0.6250 - loss: 2.7754 - val_accuracy: 0.6471 - val_loss: 2.7594
Epoch 29/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5500 - loss: 2.7661
Epoch 29: val_accuracy did not improve from 0.64706
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.5501 - loss: 2.7659 - val_accuracy: 0.5540 - val_loss: 2.7627
Epoch 30/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.5938 - loss: 2.7456
Epoch 30: val_accuracy did not improve from 0.64706
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.5938 - loss: 2.7456 - val_accuracy: 0.5882 - val_loss: 2.7517
Epoch 31/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5651 - loss: 2.7542
Epoch 31: val_accuracy did not improve from 0.64706
40/40 ━━━━━━━━━━━━━━━━━━━━ 342s 8s/step - accuracy: 0.5653 - loss: 2.7543 - val_accuracy: 0.5469 - val_loss: 2.7610
Epoch 32/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.5625 - loss: 2.7462
Epoch 32: val_accuracy did not improve from 0.64706
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.5625 - loss: 2.7462 - val_accuracy: 0.6471 - val_loss: 2.7503
Epoch 33/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5727 - loss: 2.7534
Epoch 33: val_accuracy did not improve from 0.64706
40/40 ━━━━━━━━━━━━━━━━━━━━ 341s 8s/step - accuracy: 0.5730 - loss: 2.7534 - val_accuracy: 0.5895 - val_loss: 2.7562
Epoch 34/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.6875 - loss: 2.7758
Epoch 34: val_accuracy did not improve from 0.64706
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.6875 - loss: 2.7758 - val_accuracy: 0.5294 - val_loss: 2.7355
Epoch 35/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.6055 - loss: 2.7458
Epoch 35: val_accuracy did not improve from 0.64706
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.6053 - loss: 2.7459 - val_accuracy: 0.5881 - val_loss: 2.7495
Epoch 36/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:10 6s/step - accuracy: 0.5781 - loss: 2.7696
Epoch 36: val_accuracy improved from 0.64706 to 0.67647, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 139ms/step - accuracy: 0.5781 - loss: 2.7696 - val_accuracy: 0.6765 - val_loss: 2.7163
Epoch 37/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5875 - loss: 2.7434
Epoch 37: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 344s 9s/step - accuracy: 0.5875 - loss: 2.7434 - val_accuracy: 0.5710 - val_loss: 2.7425
Epoch 38/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.6406 - loss: 2.7331
Epoch 38: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.6406 - loss: 2.7331 - val_accuracy: 0.6471 - val_loss: 2.7340
Epoch 39/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5843 - loss: 2.7414
Epoch 39: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.5843 - loss: 2.7413 - val_accuracy: 0.5795 - val_loss: 2.7380
Epoch 40/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:06 6s/step - accuracy: 0.6562 - loss: 2.7372
Epoch 40: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 105ms/step - accuracy: 0.6562 - loss: 2.7372 - val_accuracy: 0.4706 - val_loss: 2.7795
Epoch 41/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.6131 - loss: 2.7393
Epoch 41: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.6127 - loss: 2.7393 - val_accuracy: 0.5824 - val_loss: 2.7405
Epoch 42/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.5000 - loss: 2.7298
Epoch 42: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 93ms/step - accuracy: 0.5000 - loss: 2.7298 - val_accuracy: 0.3824 - val_loss: 2.7715
Epoch 43/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5961 - loss: 2.7364
Epoch 43: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5962 - loss: 2.7363 - val_accuracy: 0.5966 - val_loss: 2.7325
Epoch 44/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.6250 - loss: 2.7003
Epoch 44: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.6250 - loss: 2.7003 - val_accuracy: 0.4706 - val_loss: 2.7594
Epoch 45/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.6005 - loss: 2.7294
Epoch 45: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.6004 - loss: 2.7294 - val_accuracy: 0.6108 - val_loss: 2.7358
Epoch 46/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:08 6s/step - accuracy: 0.6250 - loss: 2.7093
Epoch 46: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.6250 - loss: 2.7093 - val_accuracy: 0.5588 - val_loss: 2.7139
Epoch 46: early stopping
Restoring model weights from the end of the best epoch: 36.
Tiempo transcurrido para el entrenamiento: 8119.172592163086 segundos
Uso de CPU durante el entrenamiento: 35.099999999999994%
Aumento en uso de memoria: 0.7940025329589844 GB
Resultados para lr=0.0005, l2=0.1, batch_size=64:
Tiempo: 8119.172592163086 segundos, CPU: 35.099999999999994%, Memoria: 0.7940025329589844 GB
Precisión de validación: 0.6764705777168274
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_40"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_20 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.2410 - loss: 2.9945
Epoch 1: val_accuracy improved from -inf to 0.46332, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 360s 2s/step - accuracy: 0.2416 - loss: 2.9939 - val_accuracy: 0.4633 - val_loss: 2.8190
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:56 2s/step - accuracy: 0.5625 - loss: 2.8076
Epoch 2: val_accuracy improved from 0.46332 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 8s 41ms/step - accuracy: 0.5625 - loss: 2.8076 - val_accuracy: 1.0000 - val_loss: 2.8018
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5013 - loss: 2.8078
Epoch 3: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 359s 2s/step - accuracy: 0.5015 - loss: 2.8077 - val_accuracy: 0.5312 - val_loss: 2.7812
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:18 2s/step - accuracy: 0.2500 - loss: 2.7819
Epoch 4: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.2500 - loss: 2.7819 - val_accuracy: 0.0000e+00 - val_loss: 2.7367
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5631 - loss: 2.7713
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.5631 - loss: 2.7712 - val_accuracy: 0.5652 - val_loss: 2.7544
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:20 2s/step - accuracy: 0.5000 - loss: 2.6942
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5000 - loss: 2.6942 - val_accuracy: 0.5000 - val_loss: 2.7598
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5797 - loss: 2.7449
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 352s 2s/step - accuracy: 0.5797 - loss: 2.7449 - val_accuracy: 0.5639 - val_loss: 2.7373
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.4375 - loss: 2.7295
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.4375 - loss: 2.7295 - val_accuracy: 1.0000 - val_loss: 2.6059
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5690 - loss: 2.7258
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 353s 2s/step - accuracy: 0.5690 - loss: 2.7258 - val_accuracy: 0.5747 - val_loss: 2.7191
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.6875 - loss: 2.7775
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6875 - loss: 2.7775 - val_accuracy: 0.5000 - val_loss: 2.5962
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5812 - loss: 2.7101
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.5811 - loss: 2.7101 - val_accuracy: 0.5707 - val_loss: 2.7060
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.6250 - loss: 2.6854
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6250 - loss: 2.6854 - val_accuracy: 1.0000 - val_loss: 2.7129
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2144.4242074489594 segundos
Uso de CPU durante el entrenamiento: 44.9%
Aumento en uso de memoria: -1.9095535278320312 GB
Resultados para lr=0.001, l2=0.01, batch_size=16:
Tiempo: 2144.4242074489594 segundos, CPU: 44.9%, Memoria: -1.9095535278320312 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_42"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_21 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.1430 - loss: 3.0686
Epoch 1: val_accuracy improved from -inf to 0.45924, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 350s 4s/step - accuracy: 0.1445 - loss: 3.0674 - val_accuracy: 0.4592 - val_loss: 2.8516
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:14 3s/step - accuracy: 0.5312 - loss: 2.8481
Epoch 2: val_accuracy improved from 0.45924 to 0.50000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 5s 29ms/step - accuracy: 0.5312 - loss: 2.8481 - val_accuracy: 0.5000 - val_loss: 2.8618
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5006 - loss: 2.8383
Epoch 3: val_accuracy improved from 0.50000 to 0.50408, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 351s 4s/step - accuracy: 0.5006 - loss: 2.8382 - val_accuracy: 0.5041 - val_loss: 2.8135
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:08 3s/step - accuracy: 0.6250 - loss: 2.7998
Epoch 4: val_accuracy did not improve from 0.50408
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6250 - loss: 2.7998 - val_accuracy: 0.5000 - val_loss: 2.7971
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5311 - loss: 2.8064
Epoch 5: val_accuracy did not improve from 0.50408
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.5311 - loss: 2.8063 - val_accuracy: 0.5027 - val_loss: 2.7916
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:11 3s/step - accuracy: 0.5312 - loss: 2.7970
Epoch 6: val_accuracy did not improve from 0.50408
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5312 - loss: 2.7970 - val_accuracy: 0.5000 - val_loss: 2.7789
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5280 - loss: 2.7826
Epoch 7: val_accuracy improved from 0.50408 to 0.53261, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 348s 4s/step - accuracy: 0.5279 - loss: 2.7825 - val_accuracy: 0.5326 - val_loss: 2.7700
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:59 3s/step - accuracy: 0.6875 - loss: 2.7776
Epoch 8: val_accuracy did not improve from 0.53261
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6875 - loss: 2.7776 - val_accuracy: 0.5000 - val_loss: 2.8851
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5491 - loss: 2.7641
Epoch 9: val_accuracy improved from 0.53261 to 0.55299, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.5492 - loss: 2.7640 - val_accuracy: 0.5530 - val_loss: 2.7579
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:02 3s/step - accuracy: 0.6562 - loss: 2.7607
Epoch 10: val_accuracy did not improve from 0.55299
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6562 - loss: 2.7607 - val_accuracy: 0.0000e+00 - val_loss: 2.7720
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5883 - loss: 2.7457
Epoch 11: val_accuracy did not improve from 0.55299
81/81 ━━━━━━━━━━━━━━━━━━━━ 344s 4s/step - accuracy: 0.5880 - loss: 2.7457 - val_accuracy: 0.5516 - val_loss: 2.7449
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:13 3s/step - accuracy: 0.6562 - loss: 2.7552
Epoch 12: val_accuracy did not improve from 0.55299
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6562 - loss: 2.7552 - val_accuracy: 0.5000 - val_loss: 2.7393
Epoch 13/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5890 - loss: 2.7356
Epoch 13: val_accuracy improved from 0.55299 to 0.55435, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.5889 - loss: 2.7356 - val_accuracy: 0.5543 - val_loss: 2.7323
Epoch 14/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:03 3s/step - accuracy: 0.5625 - loss: 2.7383
Epoch 14: val_accuracy did not improve from 0.55435
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5625 - loss: 2.7383 - val_accuracy: 0.5000 - val_loss: 2.8997
Epoch 15/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5677 - loss: 2.7240
Epoch 15: val_accuracy did not improve from 0.55435
81/81 ━━━━━━━━━━━━━━━━━━━━ 346s 4s/step - accuracy: 0.5676 - loss: 2.7240 - val_accuracy: 0.5462 - val_loss: 2.7267
Epoch 16/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.6875 - loss: 2.7204
Epoch 16: val_accuracy improved from 0.55435 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 5s 29ms/step - accuracy: 0.6875 - loss: 2.7204 - val_accuracy: 1.0000 - val_loss: 2.7205
Epoch 17/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5803 - loss: 2.7142
Epoch 17: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 353s 4s/step - accuracy: 0.5801 - loss: 2.7142 - val_accuracy: 0.5774 - val_loss: 2.7145
Epoch 18/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:17 3s/step - accuracy: 0.6250 - loss: 2.7172
Epoch 18: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.6250 - loss: 2.7172 - val_accuracy: 0.5000 - val_loss: 2.7914
Epoch 19/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5718 - loss: 2.7077
Epoch 19: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.5717 - loss: 2.7076 - val_accuracy: 0.5489 - val_loss: 2.7039
Epoch 20/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.5938 - loss: 2.6725
Epoch 20: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5938 - loss: 2.6725 - val_accuracy: 1.0000 - val_loss: 2.6707
Epoch 21/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5690 - loss: 2.7081
Epoch 21: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.5691 - loss: 2.7080 - val_accuracy: 0.5611 - val_loss: 2.7035
Epoch 22/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.4062 - loss: 2.6477
Epoch 22: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.4062 - loss: 2.6477 - val_accuracy: 1.0000 - val_loss: 2.6926
Epoch 23/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5670 - loss: 2.6927
Epoch 23: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.5669 - loss: 2.6927 - val_accuracy: 0.5367 - val_loss: 2.6929
Epoch 24/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:24 3s/step - accuracy: 0.6562 - loss: 2.6953
Epoch 24: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.6562 - loss: 2.6953 - val_accuracy: 1.0000 - val_loss: 2.8263
Epoch 25/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5528 - loss: 2.6859
Epoch 25: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.5529 - loss: 2.6859 - val_accuracy: 0.5530 - val_loss: 2.6935
Epoch 26/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:02 3s/step - accuracy: 0.5625 - loss: 2.7589
Epoch 26: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5625 - loss: 2.7589 - val_accuracy: 0.5000 - val_loss: 2.7743
Epoch 26: early stopping
Restoring model weights from the end of the best epoch: 16.
Tiempo transcurrido para el entrenamiento: 4539.860846281052 segundos
Uso de CPU durante el entrenamiento: 52.2%
Aumento en uso de memoria: 0.4283866882324219 GB
Resultados para lr=0.001, l2=0.01, batch_size=32:
Tiempo: 4539.860846281052 segundos, CPU: 52.2%, Memoria: 0.4283866882324219 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_44"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_22 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.0829 - loss: 3.1401
Epoch 1: val_accuracy improved from -inf to 0.25142, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 345s 8s/step - accuracy: 0.0842 - loss: 3.1380 - val_accuracy: 0.2514 - val_loss: 2.9238
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:53 6s/step - accuracy: 0.1406 - loss: 2.9289
Epoch 2: val_accuracy improved from 0.25142 to 0.32353, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 182ms/step - accuracy: 0.1406 - loss: 2.9289 - val_accuracy: 0.3235 - val_loss: 2.9157
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.3350 - loss: 2.8944
Epoch 3: val_accuracy improved from 0.32353 to 0.38210, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 348s 9s/step - accuracy: 0.3358 - loss: 2.8940 - val_accuracy: 0.3821 - val_loss: 2.8500
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:57 6s/step - accuracy: 0.3594 - loss: 2.8457
Epoch 4: val_accuracy improved from 0.38210 to 0.41176, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 147ms/step - accuracy: 0.3594 - loss: 2.8457 - val_accuracy: 0.4118 - val_loss: 2.8446
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4298 - loss: 2.8407
Epoch 5: val_accuracy improved from 0.41176 to 0.45597, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.4302 - loss: 2.8405 - val_accuracy: 0.4560 - val_loss: 2.8257
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.3906 - loss: 2.8352
Epoch 6: val_accuracy improved from 0.45597 to 0.50000, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 149ms/step - accuracy: 0.3906 - loss: 2.8352 - val_accuracy: 0.5000 - val_loss: 2.8217
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4776 - loss: 2.8180
Epoch 7: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.4776 - loss: 2.8179 - val_accuracy: 0.4645 - val_loss: 2.8090
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:13 6s/step - accuracy: 0.3906 - loss: 2.8047
Epoch 8: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.3906 - loss: 2.8047 - val_accuracy: 0.4412 - val_loss: 2.8013
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5102 - loss: 2.8008
Epoch 9: val_accuracy did not improve from 0.50000
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.5098 - loss: 2.8007 - val_accuracy: 0.4560 - val_loss: 2.7948
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:06 6s/step - accuracy: 0.5469 - loss: 2.7997
Epoch 10: val_accuracy improved from 0.50000 to 0.55882, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 153ms/step - accuracy: 0.5469 - loss: 2.7997 - val_accuracy: 0.5588 - val_loss: 2.7888
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4947 - loss: 2.7890
Epoch 11: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 342s 8s/step - accuracy: 0.4947 - loss: 2.7889 - val_accuracy: 0.4773 - val_loss: 2.7821
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.5156 - loss: 2.7857
Epoch 12: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 100ms/step - accuracy: 0.5156 - loss: 2.7857 - val_accuracy: 0.4412 - val_loss: 2.7594
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5052 - loss: 2.7768
Epoch 13: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5055 - loss: 2.7767 - val_accuracy: 0.4957 - val_loss: 2.7713
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.4688 - loss: 2.7815
Epoch 14: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 95ms/step - accuracy: 0.4688 - loss: 2.7815 - val_accuracy: 0.5294 - val_loss: 2.7916
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5211 - loss: 2.7668
Epoch 15: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 335s 8s/step - accuracy: 0.5211 - loss: 2.7667 - val_accuracy: 0.5142 - val_loss: 2.7636
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.5000 - loss: 2.7537
Epoch 16: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 86ms/step - accuracy: 0.5000 - loss: 2.7537 - val_accuracy: 0.5294 - val_loss: 2.7599
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5537 - loss: 2.7565
Epoch 17: val_accuracy did not improve from 0.55882
40/40 ━━━━━━━━━━━━━━━━━━━━ 344s 8s/step - accuracy: 0.5533 - loss: 2.7565 - val_accuracy: 0.5298 - val_loss: 2.7556
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.4531 - loss: 2.7510
Epoch 18: val_accuracy improved from 0.55882 to 0.61765, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 136ms/step - accuracy: 0.4531 - loss: 2.7510 - val_accuracy: 0.6176 - val_loss: 2.7415
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5439 - loss: 2.7472
Epoch 19: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5444 - loss: 2.7472 - val_accuracy: 0.5724 - val_loss: 2.7459
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:13 6s/step - accuracy: 0.5938 - loss: 2.7413
Epoch 20: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.5938 - loss: 2.7413 - val_accuracy: 0.5000 - val_loss: 2.7415
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5968 - loss: 2.7417
Epoch 21: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 335s 8s/step - accuracy: 0.5966 - loss: 2.7416 - val_accuracy: 0.5710 - val_loss: 2.7404
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:54 6s/step - accuracy: 0.5625 - loss: 2.7420
Epoch 22: val_accuracy improved from 0.61765 to 0.67647, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 158ms/step - accuracy: 0.5625 - loss: 2.7420 - val_accuracy: 0.6765 - val_loss: 2.7059
Epoch 23/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5954 - loss: 2.7328
Epoch 23: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5951 - loss: 2.7328 - val_accuracy: 0.5625 - val_loss: 2.7332
Epoch 24/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.5938 - loss: 2.7315
Epoch 24: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.5938 - loss: 2.7315 - val_accuracy: 0.5588 - val_loss: 2.7574
Epoch 25/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5745 - loss: 2.7290
Epoch 25: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.5751 - loss: 2.7289 - val_accuracy: 0.5724 - val_loss: 2.7269
Epoch 26/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:14 7s/step - accuracy: 0.5156 - loss: 2.7731
Epoch 26: val_accuracy improved from 0.67647 to 0.73529, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 165ms/step - accuracy: 0.5156 - loss: 2.7731 - val_accuracy: 0.7353 - val_loss: 2.7444
Epoch 27/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5991 - loss: 2.7192
Epoch 27: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5988 - loss: 2.7192 - val_accuracy: 0.5611 - val_loss: 2.7259
Epoch 28/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:11 6s/step - accuracy: 0.6406 - loss: 2.6833
Epoch 28: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.6406 - loss: 2.6833 - val_accuracy: 0.6176 - val_loss: 2.7019
Epoch 29/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5795 - loss: 2.7138
Epoch 29: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 345s 9s/step - accuracy: 0.5800 - loss: 2.7138 - val_accuracy: 0.5739 - val_loss: 2.7150
Epoch 30/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.5781 - loss: 2.7321
Epoch 30: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 91ms/step - accuracy: 0.5781 - loss: 2.7321 - val_accuracy: 0.5588 - val_loss: 2.7272
Epoch 31/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5716 - loss: 2.7138
Epoch 31: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.5721 - loss: 2.7136 - val_accuracy: 0.5739 - val_loss: 2.7135
Epoch 32/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:08 6s/step - accuracy: 0.7031 - loss: 2.7111
Epoch 32: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 93ms/step - accuracy: 0.7031 - loss: 2.7111 - val_accuracy: 0.5588 - val_loss: 2.7103
Epoch 33/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5900 - loss: 2.7034
Epoch 33: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 344s 9s/step - accuracy: 0.5898 - loss: 2.7034 - val_accuracy: 0.5682 - val_loss: 2.7070
Epoch 34/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:12 6s/step - accuracy: 0.5469 - loss: 2.6961
Epoch 34: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.5469 - loss: 2.6961 - val_accuracy: 0.4412 - val_loss: 2.7177
Epoch 35/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.6035 - loss: 2.6925
Epoch 35: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.6033 - loss: 2.6927 - val_accuracy: 0.5767 - val_loss: 2.7066
Epoch 36/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.6562 - loss: 2.7126
Epoch 36: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 91ms/step - accuracy: 0.6562 - loss: 2.7126 - val_accuracy: 0.5000 - val_loss: 2.7033
Epoch 36: early stopping
Restoring model weights from the end of the best epoch: 26.
Tiempo transcurrido para el entrenamiento: 6308.311935424805 segundos
Uso de CPU durante el entrenamiento: 48.3%
Aumento en uso de memoria: 0.6409111022949219 GB
Resultados para lr=0.001, l2=0.01, batch_size=64:
Tiempo: 6308.311935424805 segundos, CPU: 48.3%, Memoria: 0.6409111022949219 GB
Precisión de validación: 0.7352941036224365
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_46"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_23 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.2191 - loss: 2.9961
Epoch 1: val_accuracy improved from -inf to 0.46060, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 361s 2s/step - accuracy: 0.2198 - loss: 2.9955 - val_accuracy: 0.4606 - val_loss: 2.8177
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.3750 - loss: 2.8064
Epoch 2: val_accuracy improved from 0.46060 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 4s 16ms/step - accuracy: 0.3750 - loss: 2.8064 - val_accuracy: 1.0000 - val_loss: 2.7709
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.4738 - loss: 2.8050
Epoch 3: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.4739 - loss: 2.8049 - val_accuracy: 0.5245 - val_loss: 2.7783
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:15 2s/step - accuracy: 0.5625 - loss: 2.7841
Epoch 4: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5625 - loss: 2.7841 - val_accuracy: 0.5000 - val_loss: 2.7385
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5787 - loss: 2.7689
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.5786 - loss: 2.7689 - val_accuracy: 0.5883 - val_loss: 2.7545
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:18 2s/step - accuracy: 0.4375 - loss: 2.7306
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.4375 - loss: 2.7306 - val_accuracy: 1.0000 - val_loss: 2.6417
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5796 - loss: 2.7461
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 358s 2s/step - accuracy: 0.5795 - loss: 2.7461 - val_accuracy: 0.5652 - val_loss: 2.7348
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 0.5625 - loss: 2.7722
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5625 - loss: 2.7722 - val_accuracy: 0.0000e+00 - val_loss: 2.8733
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5674 - loss: 2.7292
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.5674 - loss: 2.7292 - val_accuracy: 0.5693 - val_loss: 2.7180
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:18 2s/step - accuracy: 0.6250 - loss: 2.7167
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6250 - loss: 2.7167 - val_accuracy: 0.5000 - val_loss: 2.9166
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5753 - loss: 2.7112
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.5753 - loss: 2.7112 - val_accuracy: 0.5788 - val_loss: 2.7057
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:10 2s/step - accuracy: 0.6875 - loss: 2.6683
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6875 - loss: 2.6683 - val_accuracy: 0.0000e+00 - val_loss: 2.6676
Epoch 12: early stopping
Restoring model weights from the end of the best epoch: 2.
Tiempo transcurrido para el entrenamiento: 2133.9292175769806 segundos
Uso de CPU durante el entrenamiento: 23.400000000000006%
Aumento en uso de memoria: -1.5922889709472656 GB
Resultados para lr=0.001, l2=0.05, batch_size=16:
Tiempo: 2133.9292175769806 segundos, CPU: 23.400000000000006%, Memoria: -1.5922889709472656 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_48"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_24 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.1535 - loss: 3.0690
Epoch 1: val_accuracy improved from -inf to 0.46332, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 346s 4s/step - accuracy: 0.1549 - loss: 3.0677 - val_accuracy: 0.4633 - val_loss: 2.8484
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.5312 - loss: 2.8454
Epoch 2: val_accuracy did not improve from 0.46332
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5312 - loss: 2.8454 - val_accuracy: 0.0000e+00 - val_loss: 2.8535
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.4983 - loss: 2.8365
Epoch 3: val_accuracy improved from 0.46332 to 0.52717, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.4985 - loss: 2.8364 - val_accuracy: 0.5272 - val_loss: 2.8099
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.5312 - loss: 2.8100
Epoch 4: val_accuracy improved from 0.52717 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 5s 25ms/step - accuracy: 0.5312 - loss: 2.8100 - val_accuracy: 1.0000 - val_loss: 2.7326
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5669 - loss: 2.8028
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 349s 4s/step - accuracy: 0.5669 - loss: 2.8027 - val_accuracy: 0.5462 - val_loss: 2.7868
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.6562 - loss: 2.7736
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6562 - loss: 2.7736 - val_accuracy: 1.0000 - val_loss: 2.8604
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5434 - loss: 2.7790
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 343s 4s/step - accuracy: 0.5434 - loss: 2.7789 - val_accuracy: 0.5652 - val_loss: 2.7702
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.4375 - loss: 2.7979
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.4375 - loss: 2.7979 - val_accuracy: 0.5000 - val_loss: 2.8309
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5694 - loss: 2.7590
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 349s 4s/step - accuracy: 0.5693 - loss: 2.7590 - val_accuracy: 0.5476 - val_loss: 2.7542
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:27 3s/step - accuracy: 0.5625 - loss: 2.7778
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 4ms/step - accuracy: 0.5625 - loss: 2.7778 - val_accuracy: 0.5000 - val_loss: 2.8150
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5744 - loss: 2.7443
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 341s 4s/step - accuracy: 0.5744 - loss: 2.7443 - val_accuracy: 0.5666 - val_loss: 2.7399
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.6250 - loss: 2.7217
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6250 - loss: 2.7217 - val_accuracy: 0.5000 - val_loss: 2.6831
Epoch 13/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5828 - loss: 2.7313
Epoch 13: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 341s 4s/step - accuracy: 0.5826 - loss: 2.7313 - val_accuracy: 0.5476 - val_loss: 2.7285
Epoch 14/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:06 3s/step - accuracy: 0.5938 - loss: 2.7162
Epoch 14: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5938 - loss: 2.7162 - val_accuracy: 0.0000e+00 - val_loss: 2.7508
Epoch 14: early stopping
Restoring model weights from the end of the best epoch: 4.
Tiempo transcurrido para el entrenamiento: 2439.8384866714478 segundos
Uso de CPU durante el entrenamiento: 58.2%
Aumento en uso de memoria: 0.38370513916015625 GB
Resultados para lr=0.001, l2=0.05, batch_size=32:
Tiempo: 2439.8384866714478 segundos, CPU: 58.2%, Memoria: 0.38370513916015625 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_50"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_25 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.1017 - loss: 3.1400
Epoch 1: val_accuracy improved from -inf to 0.25284, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 350s 8s/step - accuracy: 0.1030 - loss: 3.1379 - val_accuracy: 0.2528 - val_loss: 2.9239
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:06 6s/step - accuracy: 0.2188 - loss: 2.9276
Epoch 2: val_accuracy improved from 0.25284 to 0.26471, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 163ms/step - accuracy: 0.2188 - loss: 2.9276 - val_accuracy: 0.2647 - val_loss: 2.9167
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.3108 - loss: 2.8949
Epoch 3: val_accuracy improved from 0.26471 to 0.42188, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.3116 - loss: 2.8944 - val_accuracy: 0.4219 - val_loss: 2.8506
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:52 6s/step - accuracy: 0.2969 - loss: 2.8544
Epoch 4: val_accuracy did not improve from 0.42188
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 89ms/step - accuracy: 0.2969 - loss: 2.8544 - val_accuracy: 0.2941 - val_loss: 2.8535
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.4368 - loss: 2.8404
Epoch 5: val_accuracy improved from 0.42188 to 0.48438, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.4369 - loss: 2.8403 - val_accuracy: 0.4844 - val_loss: 2.8264
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.4531 - loss: 2.8301
Epoch 6: val_accuracy did not improve from 0.48438
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 84ms/step - accuracy: 0.4531 - loss: 2.8301 - val_accuracy: 0.4706 - val_loss: 2.8229
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4905 - loss: 2.8186
Epoch 7: val_accuracy improved from 0.48438 to 0.49148, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 345s 9s/step - accuracy: 0.4907 - loss: 2.8185 - val_accuracy: 0.4915 - val_loss: 2.8092
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:55 6s/step - accuracy: 0.4844 - loss: 2.8042
Epoch 8: val_accuracy did not improve from 0.49148
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 86ms/step - accuracy: 0.4844 - loss: 2.8042 - val_accuracy: 0.3529 - val_loss: 2.8226
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4842 - loss: 2.8043
Epoch 9: val_accuracy improved from 0.49148 to 0.53409, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 341s 8s/step - accuracy: 0.4849 - loss: 2.8042 - val_accuracy: 0.5341 - val_loss: 2.7952
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:08 6s/step - accuracy: 0.6094 - loss: 2.8036
Epoch 10: val_accuracy did not improve from 0.53409
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 93ms/step - accuracy: 0.6094 - loss: 2.8036 - val_accuracy: 0.3235 - val_loss: 2.7911
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5406 - loss: 2.7876
Epoch 11: val_accuracy improved from 0.53409 to 0.55540, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 343s 8s/step - accuracy: 0.5409 - loss: 2.7875 - val_accuracy: 0.5554 - val_loss: 2.7837
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:58 6s/step - accuracy: 0.5000 - loss: 2.7782
Epoch 12: val_accuracy improved from 0.55540 to 0.58824, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 143ms/step - accuracy: 0.5000 - loss: 2.7782 - val_accuracy: 0.5882 - val_loss: 2.7957
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5638 - loss: 2.7793
Epoch 13: val_accuracy did not improve from 0.58824
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.5639 - loss: 2.7792 - val_accuracy: 0.5483 - val_loss: 2.7736
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:03 6s/step - accuracy: 0.6250 - loss: 2.7555
Epoch 14: val_accuracy improved from 0.58824 to 0.64706, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 149ms/step - accuracy: 0.6250 - loss: 2.7555 - val_accuracy: 0.6471 - val_loss: 2.7598
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5790 - loss: 2.7669
Epoch 15: val_accuracy did not improve from 0.64706
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.5789 - loss: 2.7669 - val_accuracy: 0.5568 - val_loss: 2.7644
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:07 6s/step - accuracy: 0.5625 - loss: 2.7486
Epoch 16: val_accuracy improved from 0.64706 to 0.67647, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 13s 161ms/step - accuracy: 0.5625 - loss: 2.7486 - val_accuracy: 0.6765 - val_loss: 2.7629
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5464 - loss: 2.7587
Epoch 17: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 339s 8s/step - accuracy: 0.5469 - loss: 2.7587 - val_accuracy: 0.5554 - val_loss: 2.7559
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:01 6s/step - accuracy: 0.5938 - loss: 2.7471
Epoch 18: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.5938 - loss: 2.7471 - val_accuracy: 0.5588 - val_loss: 2.7149
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5663 - loss: 2.7485
Epoch 19: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.5661 - loss: 2.7485 - val_accuracy: 0.5398 - val_loss: 2.7512
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:17 7s/step - accuracy: 0.4688 - loss: 2.7092
Epoch 20: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.4688 - loss: 2.7092 - val_accuracy: 0.6471 - val_loss: 2.7499
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5758 - loss: 2.7409
Epoch 21: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 343s 8s/step - accuracy: 0.5761 - loss: 2.7409 - val_accuracy: 0.5966 - val_loss: 2.7392
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:12 6s/step - accuracy: 0.5781 - loss: 2.7550
Epoch 22: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 90ms/step - accuracy: 0.5781 - loss: 2.7550 - val_accuracy: 0.4706 - val_loss: 2.7560
Epoch 23/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5915 - loss: 2.7395
Epoch 23: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5918 - loss: 2.7394 - val_accuracy: 0.5724 - val_loss: 2.7343
Epoch 24/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.5312 - loss: 2.7318
Epoch 24: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 91ms/step - accuracy: 0.5312 - loss: 2.7318 - val_accuracy: 0.6176 - val_loss: 2.7428
Epoch 25/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.6086 - loss: 2.7255
Epoch 25: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.6084 - loss: 2.7256 - val_accuracy: 0.6009 - val_loss: 2.7290
Epoch 26/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:07 6s/step - accuracy: 0.5938 - loss: 2.7093
Epoch 26: val_accuracy did not improve from 0.67647
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 85ms/step - accuracy: 0.5938 - loss: 2.7093 - val_accuracy: 0.5294 - val_loss: 2.7332
Epoch 26: early stopping
Restoring model weights from the end of the best epoch: 16.
Tiempo transcurrido para el entrenamiento: 4564.721148252487 segundos
Uso de CPU durante el entrenamiento: 43.800000000000004%
Aumento en uso de memoria: 0.6228561401367188 GB
Resultados para lr=0.001, l2=0.05, batch_size=64:
Tiempo: 4564.721148252487 segundos, CPU: 43.800000000000004%, Memoria: 0.6228561401367188 GB
Precisión de validación: 0.6764705777168274
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_52"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_26 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.2899 - loss: 2.9905
Epoch 1: val_accuracy improved from -inf to 0.48098, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 355s 2s/step - accuracy: 0.2904 - loss: 2.9899 - val_accuracy: 0.4810 - val_loss: 2.8158
Epoch 2/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:23 2s/step - accuracy: 0.5000 - loss: 2.8190
Epoch 2: val_accuracy improved from 0.48098 to 0.50000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 4s 16ms/step - accuracy: 0.5000 - loss: 2.8190 - val_accuracy: 0.5000 - val_loss: 2.7869
Epoch 3/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5144 - loss: 2.8026
Epoch 3: val_accuracy improved from 0.50000 to 0.57201, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 354s 2s/step - accuracy: 0.5146 - loss: 2.8026 - val_accuracy: 0.5720 - val_loss: 2.7791
Epoch 4/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 0.5000 - loss: 2.7909
Epoch 4: val_accuracy improved from 0.57201 to 1.00000, saving model to best_model.keras
163/163 ━━━━━━━━━━━━━━━━━━━━ 4s 15ms/step - accuracy: 0.5000 - loss: 2.7909 - val_accuracy: 1.0000 - val_loss: 2.8206
Epoch 5/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5720 - loss: 2.7674
Epoch 5: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 356s 2s/step - accuracy: 0.5720 - loss: 2.7674 - val_accuracy: 0.6073 - val_loss: 2.7542
Epoch 6/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:18 2s/step - accuracy: 0.5625 - loss: 2.7952
Epoch 6: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.5625 - loss: 2.7952 - val_accuracy: 1.0000 - val_loss: 2.7404
Epoch 7/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5985 - loss: 2.7395
Epoch 7: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 358s 2s/step - accuracy: 0.5984 - loss: 2.7395 - val_accuracy: 0.5707 - val_loss: 2.7336
Epoch 8/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 3:32 1s/step - accuracy: 0.6154 - loss: 2.6923
Epoch 8: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6154 - loss: 2.6923 - val_accuracy: 0.5000 - val_loss: 2.5569
Epoch 9/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5940 - loss: 2.7259
Epoch 9: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 351s 2s/step - accuracy: 0.5939 - loss: 2.7259 - val_accuracy: 0.5693 - val_loss: 2.7175
Epoch 10/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:18 2s/step - accuracy: 0.6250 - loss: 2.7904
Epoch 10: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6250 - loss: 2.7904 - val_accuracy: 1.0000 - val_loss: 2.5673
Epoch 11/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5523 - loss: 2.7149
Epoch 11: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 350s 2s/step - accuracy: 0.5523 - loss: 2.7148 - val_accuracy: 0.5639 - val_loss: 2.7057
Epoch 12/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:08 2s/step - accuracy: 0.6250 - loss: 2.6571
Epoch 12: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6250 - loss: 2.6571 - val_accuracy: 0.5000 - val_loss: 2.6314
Epoch 13/50
163/163 ━━━━━━━━━━━━━━━━━━━━ 0s 2s/step - accuracy: 0.5554 - loss: 2.6929
Epoch 13: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 349s 2s/step - accuracy: 0.5554 - loss: 2.6929 - val_accuracy: 0.5707 - val_loss: 2.6924
Epoch 14/50
  1/163 ━━━━━━━━━━━━━━━━━━━━ 4:13 2s/step - accuracy: 0.6875 - loss: 2.5849
Epoch 14: val_accuracy did not improve from 1.00000
163/163 ━━━━━━━━━━━━━━━━━━━━ 2s 2ms/step - accuracy: 0.6875 - loss: 2.5849 - val_accuracy: 0.5000 - val_loss: 2.6987
Epoch 14: early stopping
Restoring model weights from the end of the best epoch: 4.
Tiempo transcurrido para el entrenamiento: 2492.0556807518005 segundos
Uso de CPU durante el entrenamiento: 52.4%
Aumento en uso de memoria: -0.22751235961914062 GB
Resultados para lr=0.001, l2=0.1, batch_size=16:
Tiempo: 2492.0556807518005 segundos, CPU: 52.4%, Memoria: -0.22751235961914062 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_54"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_27 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.1336 - loss: 3.0745
Epoch 1: val_accuracy improved from -inf to 0.45245, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 350s 4s/step - accuracy: 0.1350 - loss: 3.0732 - val_accuracy: 0.4524 - val_loss: 2.8566
Epoch 2/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 5:11 4s/step - accuracy: 0.5312 - loss: 2.8480
Epoch 2: val_accuracy improved from 0.45245 to 0.50000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 7s 42ms/step - accuracy: 0.5312 - loss: 2.8480 - val_accuracy: 0.5000 - val_loss: 2.8530
Epoch 3/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.4663 - loss: 2.8439
Epoch 3: val_accuracy improved from 0.50000 to 0.51902, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.4665 - loss: 2.8438 - val_accuracy: 0.5190 - val_loss: 2.8174
Epoch 4/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 3:59 3s/step - accuracy: 0.5625 - loss: 2.8044
Epoch 4: val_accuracy improved from 0.51902 to 1.00000, saving model to best_model.keras
81/81 ━━━━━━━━━━━━━━━━━━━━ 5s 27ms/step - accuracy: 0.5625 - loss: 2.8044 - val_accuracy: 1.0000 - val_loss: 2.8197
Epoch 5/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5044 - loss: 2.8104
Epoch 5: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 341s 4s/step - accuracy: 0.5048 - loss: 2.8103 - val_accuracy: 0.5448 - val_loss: 2.7917
Epoch 6/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:07 3s/step - accuracy: 0.7500 - loss: 2.7953
Epoch 6: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.7500 - loss: 2.7953 - val_accuracy: 1.0000 - val_loss: 2.7195
Epoch 7/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5741 - loss: 2.7849
Epoch 7: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 349s 4s/step - accuracy: 0.5740 - loss: 2.7848 - val_accuracy: 0.5815 - val_loss: 2.7713
Epoch 8/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:09 3s/step - accuracy: 0.5000 - loss: 2.7673
Epoch 8: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5000 - loss: 2.7673 - val_accuracy: 0.5000 - val_loss: 2.7556
Epoch 9/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5904 - loss: 2.7654
Epoch 9: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 345s 4s/step - accuracy: 0.5901 - loss: 2.7654 - val_accuracy: 0.5693 - val_loss: 2.7566
Epoch 10/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:04 3s/step - accuracy: 0.6562 - loss: 2.7819
Epoch 10: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.6562 - loss: 2.7819 - val_accuracy: 0.5000 - val_loss: 2.6378
Epoch 11/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5681 - loss: 2.7503
Epoch 11: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 342s 4s/step - accuracy: 0.5681 - loss: 2.7503 - val_accuracy: 0.5571 - val_loss: 2.7424
Epoch 12/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:37 3s/step - accuracy: 0.6562 - loss: 2.7293
Epoch 12: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 4s 5ms/step - accuracy: 0.6562 - loss: 2.7293 - val_accuracy: 0.5000 - val_loss: 2.8291
Epoch 13/50
81/81 ━━━━━━━━━━━━━━━━━━━━ 0s 3s/step - accuracy: 0.5689 - loss: 2.7411
Epoch 13: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 346s 4s/step - accuracy: 0.5688 - loss: 2.7410 - val_accuracy: 0.5598 - val_loss: 2.7334
Epoch 14/50
 1/81 ━━━━━━━━━━━━━━━━━━━━ 4:01 3s/step - accuracy: 0.5625 - loss: 2.7240
Epoch 14: val_accuracy did not improve from 1.00000
81/81 ━━━━━━━━━━━━━━━━━━━━ 3s 4ms/step - accuracy: 0.5625 - loss: 2.7240 - val_accuracy: 0.5000 - val_loss: 2.5885
Epoch 14: early stopping
Restoring model weights from the end of the best epoch: 4.
Tiempo transcurrido para el entrenamiento: 2448.003697156906 segundos
Uso de CPU durante el entrenamiento: 47.400000000000006%
Aumento en uso de memoria: -0.4161796569824219 GB
Resultados para lr=0.001, l2=0.1, batch_size=32:
Tiempo: 2448.003697156906 segundos, CPU: 47.400000000000006%, Memoria: -0.4161796569824219 GB
Precisión de validación: 1.0
Found 2621 images belonging to 18 classes.
Found 738 images belonging to 18 classes.

Model: "functional_56"

┏━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━━━━━━━━━━┳━━━━━━━━━━━━━━━┓
┃ Layer (type)                    ┃ Output Shape           ┃       Param # ┃
┡━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━━━━━━━━━━╇━━━━━━━━━━━━━━━┩
│ input_layer_28 (InputLayer)     │ (None, 224, 224, 3)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv1 (Conv2D)           │ (None, 224, 224, 64)   │         1,792 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_conv2 (Conv2D)           │ (None, 224, 224, 64)   │        36,928 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block1_pool (MaxPooling2D)      │ (None, 112, 112, 64)   │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv1 (Conv2D)           │ (None, 112, 112, 128)  │        73,856 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_conv2 (Conv2D)           │ (None, 112, 112, 128)  │       147,584 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block2_pool (MaxPooling2D)      │ (None, 56, 56, 128)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv1 (Conv2D)           │ (None, 56, 56, 256)    │       295,168 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv2 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_conv3 (Conv2D)           │ (None, 56, 56, 256)    │       590,080 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block3_pool (MaxPooling2D)      │ (None, 28, 28, 256)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv1 (Conv2D)           │ (None, 28, 28, 512)    │     1,180,160 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv2 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_conv3 (Conv2D)           │ (None, 28, 28, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block4_pool (MaxPooling2D)      │ (None, 14, 14, 512)    │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv1 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv2 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_conv3 (Conv2D)           │ (None, 14, 14, 512)    │     2,359,808 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ block5_pool (MaxPooling2D)      │ (None, 7, 7, 512)      │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ flatten (Flatten)               │ (None, 25088)          │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc1 (Dense)                     │ (None, 4096)           │   102,764,544 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ fc2 (Dense)                     │ (None, 4096)           │    16,781,312 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ predictions (Dense)             │ (None, 1000)           │     4,097,000 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_flatten (Flatten)        │ (None, 1000)           │             0 │
├─────────────────────────────────┼────────────────────────┼───────────────┤
│ custom_dense (Dense)            │ (None, 18)             │        18,018 │
└─────────────────────────────────┴────────────────────────┴───────────────┘

 Total params: 138,375,562 (527.86 MB)

 Trainable params: 18,018 (70.38 KB)

 Non-trainable params: 138,357,544 (527.79 MB)

Epoch 1/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.1474 - loss: 3.1268
Epoch 1: val_accuracy improved from -inf to 0.42756, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 355s 8s/step - accuracy: 0.1495 - loss: 3.1247 - val_accuracy: 0.4276 - val_loss: 2.9106
Epoch 2/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:11 6s/step - accuracy: 0.4219 - loss: 2.9086
Epoch 2: val_accuracy did not improve from 0.42756
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 95ms/step - accuracy: 0.4219 - loss: 2.9086 - val_accuracy: 0.3529 - val_loss: 2.9115
Epoch 3/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.4548 - loss: 2.8814
Epoch 3: val_accuracy improved from 0.42756 to 0.51705, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 344s 9s/step - accuracy: 0.4556 - loss: 2.8810 - val_accuracy: 0.5170 - val_loss: 2.8368
Epoch 4/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:14 7s/step - accuracy: 0.5312 - loss: 2.8409
Epoch 4: val_accuracy did not improve from 0.51705
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 95ms/step - accuracy: 0.5312 - loss: 2.8409 - val_accuracy: 0.4118 - val_loss: 2.8473
Epoch 5/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5584 - loss: 2.8278
Epoch 5: val_accuracy improved from 0.51705 to 0.56534, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 347s 9s/step - accuracy: 0.5584 - loss: 2.8277 - val_accuracy: 0.5653 - val_loss: 2.8125
Epoch 6/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.6094 - loss: 2.8230
Epoch 6: val_accuracy improved from 0.56534 to 0.61765, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 11s 139ms/step - accuracy: 0.6094 - loss: 2.8230 - val_accuracy: 0.6176 - val_loss: 2.8186
Epoch 7/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5597 - loss: 2.8078
Epoch 7: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 335s 8s/step - accuracy: 0.5601 - loss: 2.8077 - val_accuracy: 0.5767 - val_loss: 2.7965
Epoch 8/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.6875 - loss: 2.7892
Epoch 8: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.6875 - loss: 2.7892 - val_accuracy: 0.4706 - val_loss: 2.8048
Epoch 9/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5921 - loss: 2.7917
Epoch 9: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5920 - loss: 2.7916 - val_accuracy: 0.5682 - val_loss: 2.7835
Epoch 10/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:06 6s/step - accuracy: 0.7188 - loss: 2.7685
Epoch 10: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 84ms/step - accuracy: 0.7188 - loss: 2.7685 - val_accuracy: 0.5294 - val_loss: 2.8035
Epoch 11/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5887 - loss: 2.7812
Epoch 11: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 335s 8s/step - accuracy: 0.5886 - loss: 2.7810 - val_accuracy: 0.5795 - val_loss: 2.7718
Epoch 12/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:05 6s/step - accuracy: 0.5469 - loss: 2.7752
Epoch 12: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.5469 - loss: 2.7752 - val_accuracy: 0.6176 - val_loss: 2.7870
Epoch 13/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5894 - loss: 2.7666
Epoch 13: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 336s 8s/step - accuracy: 0.5891 - loss: 2.7665 - val_accuracy: 0.5639 - val_loss: 2.7619
Epoch 14/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:00 6s/step - accuracy: 0.5312 - loss: 2.7529
Epoch 14: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 9s 85ms/step - accuracy: 0.5312 - loss: 2.7529 - val_accuracy: 0.5000 - val_loss: 2.7929
Epoch 15/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5943 - loss: 2.7559
Epoch 15: val_accuracy did not improve from 0.61765
40/40 ━━━━━━━━━━━━━━━━━━━━ 338s 8s/step - accuracy: 0.5943 - loss: 2.7559 - val_accuracy: 0.5540 - val_loss: 2.7567
Epoch 16/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:04 6s/step - accuracy: 0.5156 - loss: 2.7750
Epoch 16: val_accuracy improved from 0.61765 to 0.64706, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 156ms/step - accuracy: 0.5156 - loss: 2.7750 - val_accuracy: 0.6471 - val_loss: 2.7465
Epoch 17/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5916 - loss: 2.7446
Epoch 17: val_accuracy did not improve from 0.64706
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5912 - loss: 2.7446 - val_accuracy: 0.5597 - val_loss: 2.7449
Epoch 18/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:09 6s/step - accuracy: 0.4688 - loss: 2.7546
Epoch 18: val_accuracy did not improve from 0.64706
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.4688 - loss: 2.7546 - val_accuracy: 0.4412 - val_loss: 2.7612
Epoch 19/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5886 - loss: 2.7400
Epoch 19: val_accuracy did not improve from 0.64706
40/40 ━━━━━━━━━━━━━━━━━━━━ 349s 9s/step - accuracy: 0.5885 - loss: 2.7399 - val_accuracy: 0.5739 - val_loss: 2.7381
Epoch 20/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:59 6s/step - accuracy: 0.5938 - loss: 2.7454
Epoch 20: val_accuracy improved from 0.64706 to 0.73529, saving model to best_model.keras
40/40 ━━━━━━━━━━━━━━━━━━━━ 12s 138ms/step - accuracy: 0.5938 - loss: 2.7454 - val_accuracy: 0.7353 - val_loss: 2.7559
Epoch 21/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5903 - loss: 2.7311
Epoch 21: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5902 - loss: 2.7311 - val_accuracy: 0.5611 - val_loss: 2.7270
Epoch 22/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:27 7s/step - accuracy: 0.5156 - loss: 2.7277
Epoch 22: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 87ms/step - accuracy: 0.5156 - loss: 2.7277 - val_accuracy: 0.5000 - val_loss: 2.7872
Epoch 23/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5701 - loss: 2.7228
Epoch 23: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 343s 8s/step - accuracy: 0.5701 - loss: 2.7228 - val_accuracy: 0.5568 - val_loss: 2.7326
Epoch 24/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:06 6s/step - accuracy: 0.6406 - loss: 2.7165
Epoch 24: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.6406 - loss: 2.7165 - val_accuracy: 0.5588 - val_loss: 2.7455
Epoch 25/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5780 - loss: 2.7185
Epoch 25: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 340s 8s/step - accuracy: 0.5780 - loss: 2.7186 - val_accuracy: 0.5653 - val_loss: 2.7243
Epoch 26/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:05 6s/step - accuracy: 0.5469 - loss: 2.7212
Epoch 26: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 88ms/step - accuracy: 0.5469 - loss: 2.7212 - val_accuracy: 0.4118 - val_loss: 2.7332
Epoch 27/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 7s/step - accuracy: 0.5562 - loss: 2.7153
Epoch 27: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 341s 8s/step - accuracy: 0.5566 - loss: 2.7153 - val_accuracy: 0.5710 - val_loss: 2.7169
Epoch 28/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 4:11 6s/step - accuracy: 0.4531 - loss: 2.7034
Epoch 28: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 89ms/step - accuracy: 0.4531 - loss: 2.7034 - val_accuracy: 0.5000 - val_loss: 2.7461
Epoch 29/50
40/40 ━━━━━━━━━━━━━━━━━━━━ 0s 6s/step - accuracy: 0.5754 - loss: 2.7116
Epoch 29: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 337s 8s/step - accuracy: 0.5753 - loss: 2.7116 - val_accuracy: 0.5440 - val_loss: 2.7082
Epoch 30/50
 1/40 ━━━━━━━━━━━━━━━━━━━━ 3:56 6s/step - accuracy: 0.5781 - loss: 2.6720
Epoch 30: val_accuracy did not improve from 0.73529
40/40 ━━━━━━━━━━━━━━━━━━━━ 10s 95ms/step - accuracy: 0.5781 - loss: 2.6720 - val_accuracy: 0.5882 - val_loss: 2.7078
Epoch 30: early stopping
Restoring model weights from the end of the best epoch: 20.
Tiempo transcurrido para el entrenamiento: 5268.132543802261 segundos
Uso de CPU durante el entrenamiento: 56.89999999999999%
Aumento en uso de memoria: 0.2124176025390625 GB
Resultados para lr=0.001, l2=0.1, batch_size=64:
Tiempo: 5268.132543802261 segundos, CPU: 56.89999999999999%, Memoria: 0.2124176025390625 GB
Precisión de validación: 0.7352941036224365
Mejores hiperparámetros encontrados:
{'learning_rate': 0.0001, 'l2_regularization': 0.01, 'batch_size': 16, 'val_accuracy': 1.0, 'elapsed_time': 4630.410501003265, 'cpu_usage': 46.2, 'memory_usage': -264822784}

graficas de entrenamiento y validación

import matplotlib.pyplot as plt

def plotTraining(hist):
    epochs = range(len(hist.history['loss']))
    
    plt.figure(figsize=(12, 10))

    # Gráfico de la pérdida
    plt.subplot(2, 1, 1)
    plt.plot(epochs, hist.history['loss'], '-r', label='Pérdida del entrenamiento')
    plt.plot(epochs, hist.history['val_loss'], '-b', label='Pérdida de validación')
    plt.title('Pérdida del entrenamiento y validación')
    plt.xlabel('Épocas', fontsize=14)
    plt.ylabel('Pérdida', fontsize=14)
    plt.legend()
    plt.grid()

    # Gráfico de la precisión
    plt.subplot(2, 1, 2)
    plt.plot(epochs, hist.history['accuracy'], '-r', label='Precisión del entrenamiento')
    plt.plot(epochs, hist.history['val_accuracy'], '-b', label='Precisión de validación')
    plt.title('Precisión del entrenamiento y validación')
    plt.xlabel('Épocas', fontsize=14)
    plt.ylabel('Precisión', fontsize=14)
    plt.legend()
    plt.grid()

    plt.tight_layout()
    plt.show()

# Uso de la función para graficar
plotTraining(model_history)

from sklearn.metrics import confusion_matrix, f1_score, roc_curve, precision_score, recall_score, accuracy_score, roc_auc_score
from sklearn import metrics
from mlxtend.plotting import plot_confusion_matrix
from keras.models import load_model
import numpy as np
import matplotlib.pyplot as plt
%matplotlib inline


names = [ 'CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS', 'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS', 'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA', 'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']



test_data_dir = 'datasetpreprocesado/test'  

test_datagen = ImageDataGenerator()

test_generator = test_datagen.flow_from_directory(
    test_data_dir,
    target_size=(width_shape, height_shape), 
    batch_size = batch_size,
    class_mode='categorical', 
    shuffle=False)

custom_Model50= load_model('best_model.keras')

predictions = custom_Model50.predict(test_generator)


y_pred = np.argmax(predictions, axis=1)
y_real = test_generator.classes


matc=confusion_matrix(y_real, y_pred)

plot_confusion_matrix(conf_mat=matc, figsize=(9,9), class_names = names, show_normed=False)
plt.tight_layout()

print(metrics.classification_report(y_real,y_pred, digits = 4))

Found 361 images belonging to 18 classes.

C:\Users\Oscar Diaz\anaconda3\Lib\site-packages\keras\src\trainers\data_adapters\py_dataset_adapter.py:120: UserWarning: Your `PyDataset` class should call `super().__init__(**kwargs)` in its constructor. `**kwargs` can include `workers`, `use_multiprocessing`, `max_queue_size`. Do not pass these arguments to `fit()`, as they will be ignored.
  self._warn_if_super_not_called()

6/6 ━━━━━━━━━━━━━━━━━━━━ 42s 6s/step
              precision    recall  f1-score   support

           0     0.4783    0.5500    0.5116        20
           1     0.4444    0.2000    0.2759        20
           2     0.6000    0.4500    0.5143        20
           3     0.6667    0.1000    0.1739        20
           4     0.2812    0.4500    0.3462        20
           5     0.1509    0.4000    0.2192        20
           6     0.4872    0.9500    0.6441        20
           7     0.1967    0.6000    0.2963        20
           8     0.1111    0.0500    0.0690        20
           9     0.0385    0.0500    0.0435        20
          10     0.2500    0.0500    0.0833        20
          11     0.2381    0.2500    0.2439        20
          12     0.1250    0.0500    0.0714        20
          13     0.0952    0.1000    0.0976        20
          14     0.7778    0.7000    0.7368        20
          15     0.0000    0.0000    0.0000        20
          16     0.3125    0.2500    0.2778        20
          17     0.5000    0.0476    0.0870        21

    accuracy                         0.2909       361
   macro avg     0.3196    0.2915    0.2606       361
weighted avg     0.3201    0.2909    0.2602       361

from keras.applications.imagenet_utils import preprocess_input, decode_predictions
from keras.models import load_model
import cv2

names =  [
    'CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 
    'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS',
    'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS',
    'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA',
    'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS']

# Cargar el modelo
modelt = load_model("models\entrenamiento.keras")
#modelt = custom_vgg_model

# Ruta de la imagen de prueba
imaget_path = 'output\CYANOCORAX YNCAS_CYANOCORAX YNCAS_1.jpg'

# Leer la imagen, cambiar tamaño y preprocesar
imaget=cv2.resize(cv2.imread(imaget_path), (width_shape, height_shape), interpolation = cv2.INTER_AREA)
xt = np.asarray(imaget)
xt=preprocess_input(xt)
xt = np.expand_dims(xt,axis=0)

# Obtener las predicciones del modelo
preds = modelt.predict(xt)

# Obtener la clase predicha y su porcentaje de confianza
predicted_class_index = np.argmax(preds)
predicted_class_name = names[predicted_class_index]
confidence_percentage = preds[0][predicted_class_index] * 100

# Imprimir el resultado
print(f'Clase predicha: {predicted_class_name}')
print(f'Porcentaje de confianza: {confidence_percentage:.2f}%')

# Mostrar la imagen
plt.imshow(cv2.cvtColor(np.asarray(imaget), cv2.COLOR_BGR2RGB))
plt.axis('off')
plt.show()

1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 991ms/step
Clase predicha: CYANOCORAX YNCAS
Porcentaje de confianza: 98.90%

import os
import cv2
from keras.applications.imagenet_utils import preprocess_input
from keras.models import load_model
import numpy as np
import matplotlib.pyplot as plt

# Definir el tamaño de las imágenes
width_shape = 224
height_shape = 224

# Nombres de las clases
class_names = [
    'CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 
    'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS',
    'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS',
    'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA',
    'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS'
]

# Cargar el modelo
model = load_model("models/entrenamiento.keras")

# Directorio de salida
output_dir = "output"

# Obtener una lista de todas las imágenes en el directorio de salida
image_files = [f for f in os.listdir(output_dir) if os.path.isfile(os.path.join(output_dir, f))]

# Número de columnas por fila
num_cols = 5

# Número de imágenes procesadas
num_processed = 0

# Crear una nueva figura
plt.figure(figsize=(20, 20))

# Iterar sobre cada imagen
for image_file in image_files:
    # Ruta completa de la imagen
    image_path = os.path.join(output_dir, image_file)
    
    # Leer la imagen, cambiar tamaño y preprocesar
    image = cv2.imread(image_path)
    image = cv2.resize(image, (width_shape, height_shape), interpolation=cv2.INTER_AREA)
    image = np.asarray(image)
    image = preprocess_input(image)
    image = np.expand_dims(image, axis=0)

    # Obtener las predicciones del modelo
    preds = model.predict(image)

    # Obtener la clase predicha y su porcentaje de confianza
    predicted_class_index = np.argmax(preds)
    confidence_percentage = preds[0][predicted_class_index] * 100
    predicted_class_name = class_names[predicted_class_index]

    # Mostrar la imagen y la predicción
    plt.subplot((len(image_files) - 1) // num_cols + 1, num_cols, num_processed + 1)
    plt.imshow(cv2.cvtColor(cv2.imread(image_path), cv2.COLOR_BGR2RGB))
    plt.title(f'Clase: {predicted_class_index} - {predicted_class_name}\nNombre: {image_file}\nConfianza: {confidence_percentage:.2f}%')
    plt.axis('off')
    
    # Incrementar el contador de imágenes procesadas
    num_processed += 1

# Ajustar el espacio entre subtramas y mostrar la figura
plt.tight_layout()
plt.show()

1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 839ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 636ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 660ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 649ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 650ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 770ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 850ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 830ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 909ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 882ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 957ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 994ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 975ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 835ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 973ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 876ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 861ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 875ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 904ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 890ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 880ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 841ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 930ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 865ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 980ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 2s 2s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 910ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 875ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 850ms/step

import os
import cv2
from keras.applications.imagenet_utils import preprocess_input
from keras.models import load_model
import numpy as np
import matplotlib.pyplot as plt

# Definir el tamaño de las imágenes
width_shape = 224
height_shape = 224

# Nombres de las clases
class_names = [
    'CATHARTES AURA', 'COEREBA FLAVEOLA', 'COLUMBA LIVIA', 'CORAGYPS ATRATUS', 
    'CROTOPHAGA SULCIROSTRIS', 'CYANOCORAX YNCAS', 'EGRETTA THULA', 'FALCO PEREGRINUS',
    'FALCO SPARVERIUS', 'HIRUNDO RUSTICA', 'PANDION HALIAETUS', 'PILHERODIUS PILEATUS',
    'PITANGUS SULPHURATUS', 'PYRRHOMYIAS CINNAMOMEUS', 'RYNCHOPS NIGER', 'SETOPHAGA FUSCA',
    'SYNALLAXIS AZARAE', 'TYRANNUS MELANCHOLICUS'
]

# Cargar el modelo
model = load_model("models/optimizado.keras")

# Directorio de salida
output_dir = "output"

# Obtener una lista de todas las imágenes en el directorio de salida
image_files = [f for f in os.listdir(output_dir) if os.path.isfile(os.path.join(output_dir, f))]

# Número de columnas por fila
num_cols = 5

# Número de imágenes procesadas
num_processed = 0

# Crear una nueva figura
plt.figure(figsize=(20, 20))

# Iterar sobre cada imagen
for image_file in image_files:
    # Ruta completa de la imagen
    image_path = os.path.join(output_dir, image_file)
    
    # Leer la imagen, cambiar tamaño y preprocesar
    image = cv2.imread(image_path)
    image = cv2.resize(image, (width_shape, height_shape), interpolation=cv2.INTER_AREA)
    image = np.asarray(image)
    image = preprocess_input(image)
    image = np.expand_dims(image, axis=0)

    # Obtener las predicciones del modelo
    preds = model.predict(image)

    # Obtener la clase predicha y su porcentaje de confianza
    predicted_class_index = np.argmax(preds)
    confidence_percentage = preds[0][predicted_class_index] * 100
    predicted_class_name = class_names[predicted_class_index]

    # Mostrar la imagen y la predicción
    plt.subplot((len(image_files) - 1) // num_cols + 1, num_cols, num_processed + 1)
    plt.imshow(cv2.cvtColor(cv2.imread(image_path), cv2.COLOR_BGR2RGB))
    plt.title(f'Clase: {predicted_class_index} - {predicted_class_name}\nNombre: {image_file}\nConfianza: {confidence_percentage:.2f}%')
    plt.axis('off')
    
    # Incrementar el contador de imágenes procesadas
    num_processed += 1

# Ajustar el espacio entre subtramas y mostrar la figura
plt.tight_layout()
plt.show()

1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 848ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 621ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 626ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 629ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 621ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 670ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 679ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 663ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 804ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 868ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 756ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 875ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 848ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 809ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 819ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 956ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step    
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 781ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 803ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 807ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 804ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 825ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 995ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 960ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 957ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 958ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 1s/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 894ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 822ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 848ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 797ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 838ms/step
1/1 ━━━━━━━━━━━━━━━━━━━━ 1s 888ms/step

from tensorflow.keras.models import load_model
from tensorflow.keras.utils import plot_model

# Cargar el modelo
model = load_model("models/optimizado.keras")

# Guardar la representación gráfica de la arquitectura del modelo en un archivo
plot_model(model, to_file='model_plot.png', show_shapes=True, show_layer_names=True)

C:\Users\DIAZOVIEDO\anaconda3\envs\TFMaves\lib\site-packages\keras\src\saving\saving_lib.py:396: UserWarning: Skipping variable loading for optimizer 'adam', because it has 58 variables whereas the saved optimizer has 6 variables. 
  trackable.load_own_variables(weights_store.get(inner_path))

from tensorflow.keras.models import load_model
from tensorflow.keras.utils import plot_model

# Cargar el modelo
model = load_model("models/optimizado.keras")

# Generar un gráfico más compacto
plot_model(model, to_file='model_plot.png', show_shapes=False, show_layer_names=False)

C:\Users\DIAZOVIEDO\anaconda3\envs\TFMaves\lib\site-packages\keras\src\saving\saving_lib.py:396: UserWarning: Skipping variable loading for optimizer 'adam', because it has 58 variables whereas the saved optimizer has 6 variables. 
  trackable.load_own_variables(weights_store.get(inner_path))

import tensorflow as tf
from tensorflow.keras.models import load_model
from IPython.display import Image
from graphviz import Digraph

# Cargar modelos optimizados y entrenados
model_entrenamiento = load_model("models/entrenamiento.keras")
model_optimizado = load_model("models/optimizado.keras")

# Generar diagramas de flujo para cada modelo
def generar_diagrama_flujo(modelo, nombre_modelo):
  """
  Genera un diagrama de flujo de la arquitectura del modelo y lo muestra en Jupyter Notebook.

  Args:
    modelo: Instancia del modelo Keras.
    nombre_modelo: Nombre del modelo para identificar el diagrama.
  """
  
  # Convertir la arquitectura del modelo a formato Graphviz
  dot_graph = tf.keras.utils.model_to_dot(modelo, show_shapes=True)

  # Crear un objeto Digraph a partir del código Graphviz
  graph = Digraph(nombre_modelo)
  graph.from_dot(dot_graph)

  # Personalizar el estilo del diagrama (opcional)
  graph.node_attr['shape'] = 'box'
  graph.edge_attr['style'] = 'arrowsize=2'

  # Mostrar el diagrama en Jupyter Notebook
  Image(graph)

# Generar diagrama de flujo para el modelo de entrenamiento
generar_diagrama_flujo(model_entrenamiento, "Modelo entrenamiento")

# Generar diagrama de flujo para el modelo optimizado
generar_diagrama_flujo(model_optimizado, "Modelo optimizado")

C:\Users\DIAZOVIEDO\anaconda3\envs\TFMaves\lib\site-packages\keras\src\saving\saving_lib.py:396: UserWarning: Skipping variable loading for optimizer 'adam', because it has 66 variables whereas the saved optimizer has 6 variables. 
  trackable.load_own_variables(weights_store.get(inner_path))

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
Cell In[13], line 35
     32   Image(graph)
     34 # Generar diagrama de flujo para el modelo de entrenamiento
---> 35 generar_diagrama_flujo(model_entrenamiento, "Modelo entrenamiento")
     37 # Generar diagrama de flujo para el modelo optimizado
     38 generar_diagrama_flujo(model_optimizado, "Modelo optimizado")

Cell In[13], line 25, in generar_diagrama_flujo(modelo, nombre_modelo)
     23 # Crear un objeto Digraph a partir del código Graphviz
     24 graph = Digraph(nombre_modelo)
---> 25 graph.from_dot(dot_graph)
     27 # Personalizar el estilo del diagrama (opcional)
     28 graph.node_attr['shape'] = 'box'

AttributeError: 'Digraph' object has no attribute 'from_dot'

from tensorflow.keras.models import load_model

# Cargar el modelo
model = load_model("models/optimizado.keras")

# Exportar el modelo a formato ONNX
model.save('model.onnx')

# Abrir Netron y cargar el modelo ONNX
netron model.onnx

  Cell In[8], line 10
    netron model.onnx
           ^
SyntaxError: invalid syntax

from tensorflow.keras.models import load_model
from visdom import Visdom

# Cargar el modelo
model = load_model("models/optimizado.keras")

# Crear un objeto Visdom
vis = Visdom()

# Visualizar el modelo
vis.modelsummary(model, name='model')

---------------------------------------------------------------------------
ModuleNotFoundError                       Traceback (most recent call last)
Cell In[4], line 2
      1 from tensorflow.keras.models import load_model
----> 2 from visdom import Visdom
      4 # Cargar el modelo
      5 model = load_model("models/optimizado.keras")

ModuleNotFoundError: No module named 'visdom'